Warm-blood cardioplegia with low or high magnesium for coronary bypass surgery: a randomised controlled trial.

OBJECTIVE: Magnesium (Mg²⁺) is cardioprotective and has been routinely used to supplement cardioplegic solutions during coronary artery bypass graft (CABG) surgery. However, there is no consensus about the Mg²⁺ concentration that should be used. The aim of this study was to compare the effects of intermittent antegrade warm-blood cardioplegia supplemented with either low- or high-concentration Mg²⁺.
METHODS: This study was a randomised controlled trial carried out in two cardiac surgery centres, Bristol, UK and Cuneo, Italy. Patients undergoing isolated CABG with cardiopulmonary bypass were eligible. Patients were randomised to receive warm-blood cardioplegia supplemented with 5 or 16 mmol l⁻¹ Mg². The primary outcome was postoperative atrial fibrillation. Secondary outcomes were serum biochemical markers (troponin I, Mg²⁺, potassium, lactate and creatinine) and time-to-plegia arrest. Intra-operative and postoperative clinical outcomes were also recorded.
RESULTS: Data from two centres for 691 patients (342 low and 349 high Mg²⁺) were analysed. Baseline characteristics were similar for both groups. There was no significant difference in the frequency of postoperative atrial fibrillation in the high (32.8%) and low (32.0%) groups (risk ratio 1.03, 95% confidence interval, CI, 0.82-1.28). However, compared with the low group, troponin I release was 28% less (95% CI 55-94%, p=0.02) in the high-Mg²⁺ group. The 30-day mortality was 0.72% (n = 5); all deaths occurred in the high-Mg²⁺ group but there was no significant difference between the groups (p=0.06). Frequencies of other major complications were similar in the two groups.
CONCLUSIONS: Warm-blood cardioplegia supplemented with 16 mmol l⁻¹ Mg²⁺, compared with 5 mmol l⁻¹ Mg²⁺, does not reduce the frequency of postoperative atrial fibrillation in patients undergoing CABG but may reduce cardiac injury. (This trial was registered as ISRCTN95530505.).

RCT Entities:   Population Interventions Outcomes

Introduction

Researchers have investigated the cardioprotective effects of magnesium (Mg2+) for over 30 years. A Cochrane review of intravenous Mg2+ after myocardial infarction (MI) concluded that Mg2+ is unlikely to reduce mortality but may reduce the incidence of arrhythmia [1]. Another, recent systematic review of intravenous Mg2+ after coronary artery bypass graft (CABG) surgery concluded that Mg2+ prevents postoperative atrial fibrillation (POAF) [2].

Mg2+ may also exert a cardioprotective effect during ischaemia and reperfusion through a beneficial effect on calcium (Ca2+) transport. Hyperkalaemic cardioplegic solutions partially depolarise the membrane and may open the L-type Ca2+ channels. Elevated intracellular Ca2+ levels activate a variety of cellular enzymes and transport systems as well as influencing mitochondrial function and increasing cellular energy demands [3-5]. Mg2+ blocks the L-type Ca2+ channels, reduces Ca2+ loading and energy demands and preserves myocardial metabolites. We have previously shown that adding Mg2+ to intermittent warm-blood cardioplegia minimised the ischaemic metabolic derangement and reduced the reperfusion-induced injury observed in the hearts of patients undergoing elective CABG [6].

In a previous randomised trial, we demonstrated that adding a relatively low amount of Mg2+ (5 mmol l−1) to intermittent antegrade warm-blood cardioplegia (IAWBC) resulted in some benefit in patients undergoing urgent CABG for unstable angina [7]. We hypothesised that this benefit might have been greater if the Mg2+ concentration had been higher (≈16 mmol l−1). This concentration has been shown to confer optimal protection to rat heart perfused with crystalloid cardioplegic solution [8].

Thus, the objective of the present randomised controlled trial was to compare the effect of IAWBC supplemented with high- (16 mmol l−1) or low- (5 mmol l−1) dose Mg2+ on the risk of POAF in patients undergoing isolated CABG.

Materials and methods

Study population

All adult patients undergoing isolated CABG with cardiopulmonary bypass (CPB) were candidates for the trial. Patients were excluded if they required a repeat operation, were in cardiogenic shock or had heart valve pathology and renal or hepatic dysfunction.

Participants were randomly assigned in a 1:1 ratio to parallel groups, that is, IAWBC supplemented with Mg2+ at a concentration of 5 mmol l−1 (‘low’ group) or 16 mmol l−1 (‘high’ group). Two centres (Bristol, UK and Cuneo, Italy) participated. The study was approved by the Central and South Bristol National Health Services (NHS) Research Ethics Committee (reference E5183). All patients gave written informed consent.

Anaesthetic, surgical and postoperative care

The standardised anaesthetic technique consisted of oral premedication (20 mg temazepam), induction with 2 mg kg−1 propofol followed by total intravenous anaesthesia with propofol (3 mg kg−1 h−1). Neuromuscular blockade was achieved by administering pancuronium bromide (0.15 mg kg−1) or vecuronium (0.15 mg kg−1). Intravenous heparin (300 IU kg−1) was administered immediately before cannulation for CPB and additional doses were given to maintain an activated clotting time (ACT) of 480 s or greater.

IAWBC is used routinely in both centres because of evidence of better myocardial protection [6,9,10]. Blood cardioplegia was prepared by mixing whole blood with potassium chloride (KCl) and magnesium sulphate (MgSO4) solution to create two different concentrations of Mg2+ (5 and 16 mmol l−1), using the methods described by Calafiore and colleagues [10].

CPB was instituted by cannulation of the distal ascending aorta and insertion of a single two-stage cannula into the right atrium. Blood was withdrawn directly from the pump oxygenator and then reinfused at 34–37 °C into the aortic root using a roller pump. Non-pulsatile flow rates of 2.4 l min−1 m−2 were used, as previously described [7]. The first dose lasted 2 min at a flow rate of 300 ml min−1 (600 ml overall). The time to cardioplegic arrest was recorded. The syringe pump delivered 2 ml of the solution in approximately 20 s, and then the flow rate was reduced to 150 ml min−1, achieving a final KCl concentration of 20 mmol l−1 in both groups and an MgSO4 concentration of 5 or 16 mmol l−1, respectively. After each distal anastomosis (or after 15 min of ischaemia), a second dose of cardioplegia was administered at a flow rate of 200 ml min−1 (blood) and 120 ml h−1 (KCl/MgSO4 solutions) for a further 2 min. Postoperative management (fluid balance, transfusion requirements and inotropic support, etc.) was in accordance with the unit protocols previously described [11].

Distal anastomoses were completed first during a single period of aortic cross-clamping. Proximal anastomoses were performed following removal of the aortic cross-clamp and on the beating heart using an aortic side-biting clamp. At the end of the surgical procedure, protamine sulphate was administered to reverse the effect of heparin.

Randomisation

Random treatment allocations, stratified by centre and blocked to ensure approximate balance in the number of patients allocated to each group, were generated by computer before starting the study. One of the authors (BCR) concealed the allocations in sequentially numbered, sealed, opaque envelopes, which were kept in theatre. Trial participants were recruited by the surgical team. The perfusionist, the only member of the care team who was not blind to allocation, opened the next numbered envelope immediately before starting the operation, and was responsible for preparing the cardioplegia solution in accordance with the allocation. The trial is registered as ISRCTN95530505 with Current Controlled Trials.

Outcomes, data collection and definitions

The prespecified primary outcome was new POAF during the index hospital admission. Secondary outcomes were serum biochemical markers (troponin I, Mg2+, potassium, lactate and creatinine) and time-to-plegia arrest. A range of other intra-operative and postoperative clinical outcomes during hospital admission were also recorded.

Baseline characteristics, intra-operative details and postoperative clinical outcomes were collected for all participants. Operative priority was defined as urgent, if the patient had unstable angina requiring in-hospital treatment. Heart rate and rhythm were continuously monitored during the first 72 h after surgery (equivalent to using a Holter electrocardiograph (ECG) tape). Twelve-lead ECG recordings were performed before surgery, 2 h after surgery and then daily until hospital discharge. An ECG was also recorded on the basis of any clinical suspicion of arrhythmia. When arrhythmia was documented, ECG monitoring was restarted. Each episode of arrhythmia was interpreted by an intensive care physician blinded to treatment allocation, who made the diagnosis of POAF. Atrial fibrillation, atrial flutter and atrial tachycardia were defined according to Kalman and colleagues [12]. Indications for temporary pacing were also assessed by intensive care physicians and included symptomatic bradycardia unresponsive to treatment, acute conduction disturbances including second- or third-degree atrioventricular block and bifascicular or trifascicular block.

We planned to measure serum Mg2+, cardiac troponin I, potassium, lactate and creatinine concentrations at baseline, 4, 12, 24 and 48 h after surgery in a consecutive subsample of participants. A subsample was considered adequate because continuously scaled outcomes with repeated measures provide more power. Surrogate clinical markers of myocardial preservation were also assessed, including the need for defibrillation on removal of the aortic cross-clamp, post-CPB inotropic agents or use of an intra-aortic balloon pump. Operative mortality was defined as any death within 30 days of operation. Other major complications were defined and recorded as previously [11].

Statistical power and data analyses

The risk of new POAF in the study population was estimated to be 25% (95% confidence interval, CI, 21–30%) [7], a frequency similar to that reported recently [2]. We assumed that a 10% reduction in this frequency (equivalent to a risk ratio of 0.6) would be clinically important, requiring a sample size of 354 patients in each group for the study to have a 90% chance of detecting a difference of this size or greater between groups at a 5% significance level (two-tailed). We aimed to randomise 800 participants in total, to allow for some loss to follow-up. Because biomarkers were continuously scaled and measured several times after surgery, a much smaller sample size was required for these outcomes. For example, based on the standard deviation for logarithmically transformed postoperative troponin data from a previous trial [13], a sample size of 72 was required to detect a difference of 0.3 between the troponin geometric means between groups, equivalent to an overall reduction of ≥26% (assuming no interaction, 90% power and a 5% two-tailed significance level).

All comparisons between groups were carried out according to participants’ original allocations but without interpolating missing outcome data. Categorical outcomes were compared by chi-squared tests, estimating risk ratios with 95% CIs. Times to plegia arrest and biochemical data were positively skewed and transformed logarithmically; hence, geometric means and ratios of geometric means are reported. Time-to-plegia arrest was compared between groups by a t-test. Biochemical markers were compared using mixed regression models to take account of the repeated measures, adjusting for preoperative measurements of the markers and participating centre. Interactions of treatment effect with time were tested (F-tests) and, if statistically significant at the 5% level, ratios were reported separately for each time point; otherwise, a single ratio estimate of the overall effect was reported.

Two post hoc analyses were carried out. One analysis used logistic regression to carry out a subgroup analysis by operative priority (elective vs urgent). The second compared survival between groups up to 5 years using the Kaplan–Meier log-rank test.

Reporting

This trial is reported in accordance with international recommendations for the reporting of randomised trials (http://www.consort-statement.org/consort-statement/overview0/, accessed 9 April 2010) [14].

Results

Participants were recruited between 14 January 2002 and 10 July 2007 (741 in total, 566 in Bristol and 175 in Cuneo). The trial stopped short of its recruitment target because of a lack of resources. The first 40 participants from Bristol (up to 22 May 2002) were excluded from all analyses because data quality was extremely poor during this period (Fig. 1). A further 10 participants in Bristol were excluded because the Mg2+ level received was not recorded. These exclusions left 691 participants for analysis, 342 of whom were randomised to the low group and 349 to the high group; 24 received the wrong Mg2+ concentration.

Fig. 1

Flow diagram showing the numbers of participants randomised in the trial and with data contributing to analyses of different outcomes. CPB: cardiopulmonary bypass; MI: myocardial infarction; and IABC: intra-aortic balloon pump. For numbers of participants contributing data for serum biomarkers, see Table 4.

The characteristics of participants at baseline were balanced across groups and are summarised in Table 1. Operative details, that is, ECG ischaemia before CPB (3%), CPB and cross-clamp times (mean and standard deviation 41 ± 12 min and 75 ± 20 min, respectively), number of distal anastomoses (77% of participants had ≥3 distal anastomoses) and ECG ischaemia after CPB (5%) were also balanced across groups.

Table 1

Characteristics of participants in 5 mmol l−1 Mg2+ (n = 342) and 16 mmol l−1 Mg2+ (n = 349) groups at baseline.

	Concentration of Mg2+ in cardioplegia
	5 mmol l−1 (n = 342)		16 mmol l−1 (n = 349)
	Frequency	%	Frequency	%
Age (years)a	66.0	9.36	65.5	8.48
Female	64	18.7%	54	15.5%
Urgent (cf. elective) priority	101	31.9%	93	28.3%
EuroSCOREb	3	1–5	3	1–4
CCS class
0	2	0.7%	1	0.3%
I	49	16.3%	47	14.6%
II	116	38.5%	127	39.4%
III	77	25.6%	96	29.8%
IV	57	18.9%	51	15.8%
NYHA class
0	2	0.7%	0	0
I	81	27.9%	85	27.5%
II	131	45.2%	161	52.1%
III	68	23.5%	57	18.5%
IV	8	2.8%	6	1.9%
Unstable angina	94	31.0%	90	28%
Congestive cardiac failure	18	8%	18	7.6%
Ejection fraction
Good (>49%)	257	78.8%	261	77.7%
Fair (30–49%)	61	18.7%	64	19.1%
Poor (<30%)	8	2.5%	11	3.3%
Previous myocardial infarction	145	44.3%	153	45.5%
Family history of IHD	116	39.3%	137	44.5%
Hypertension	241	74.4%	246	73.2%
Smoking status
Never smoked
Ex smoker	165	53.2%	162	50.6%
Current smoker	45	14.5%	45	14.1%
Diabetic status
None
Diet controlled	7	2.2%	3	1%
NIDDM	47	15.0%	38	11.7%
IDDM	16	5.1%	24	7.4%
Chronic obstructive airways disease	31	9.5%	26	7.8%
Neurological disease (CVA/TIA)	23	7.1%	20	6.1%
Peripheral vascular disease	33	10%	32	9.5%
Preoperative atrial fibrillation (cf sinus/other rhythm)	14	4.3%	10	3.0%

CCS: Canadian Cardiovascular Society; NYHA: New York Heart Association; IHD: ischaemic heart disease; NIDDM: non-insulin dependent diabetes mellitus; IDDM: insulin dependent diabetes mellitus; and CVA/TIA: cerebrovascular accident/transient ischaemic attack.

Mean and standard deviation.

Median and interquartile range.

New postoperative atrial fibrillation and other clinical outcomes

Findings for the primary and secondary clinical outcomes are summarised in Table 2. There was no difference in the frequency of new POAF between groups (low, 32.0%; high, 32.8%; risk ratio 1.03, 95% CI 0.82–1.28, p = 0.82). Prompted by the previous trial result [7], we did a post hoc subgroup analysis comparing elective and urgent patients (restricted to 613 patients for whom operative priority was known). The percentage of participants having an urgent operation was similar in the high and low groups (Table 1). The test of interaction was marginally significant (p = 0.05, Table 3). Among elective participants, high-Mg2+ concentration slightly increased the odds of POAF (odds ratio (OR) = 1.31, 95% CI 0.86–1.99). Among urgent participants, high concentration reduced the odds of POAF (OR = 1.31 × 0.48 = 0.63, 95% CI 0.35–1.13). In terms of operative priority, urgent priority increased the odds of POAF (OR = 2.34, 95% CI 1.42–3.86) in the low group but not in the high group (OR = 2.34 × 0.48 = 1.12, 95% CI 0.68–1.88).

Table 2

Primary and secondary outcomes for the 5 mmol l−1 and 16 mmol l−1 Mg2+ groups.

	Low group (n = 342)		High group (n = 349)		High versus low groups		pa
	Frequency	%	Frequency	%	Risk ratio	95% CI
New postoperative atrial fibrillation	102	32.0%	107	32.8%	1.03	0.82–1.28	0.82
New postoperative arrhythmia					na	na	0.37
None	209	65.5%	213	65.3%
Atrial fibrillation	102	32.0%	107	32.8%
Other	8	2.5%	6	1.8%
Defibrillation after CPB	27	8.7%	42	13.0%	1.50	0.95–2.37	0.08
Postoperative DC shock	16	5.2%	12	3.8%	0.73	0.35–1.51	0.39
Postoperative pacing	17	5.5%	14	4.4%	0.79	0.40–1.58	0.51
Postoperative myocardial infarction	4	1.2%	3	0.9%	0.74	0.17–3.27	0.69
Postoperative IABP	0	0.0%	3	0.9%	na	na	0.13
Postoperative inotropes (any)	90	28.5%	85	26.4%	0.93	0.72–1.19	0.56
Postoperative inotropes (dose)b					na	na	0.91
None	226	71.5%	237	73.6%
Minimum	68	21.5%	63	19.6%
Moderate	18	5.7%	17	5.3%
Maximum	4	1.3%	5	1.6%

CPB: cardiopulmonary bypass; IABC: intra-aortic balloon pump; and na: risk ratio not calculable because the outcome did not occur in one or other group. Some outcomes for some patients were missing (see Fig. 1).

p value based on chi-squared test or Fisher exact test for low frequencies.

Inotropic requirement: None ≤ 3 μg kg−1 min−1 dopamine, minimum = 3–5 μg kg−1 min−1 dopamine, Moderate ≥ 5 μg kg−1 min−1, Maximum = adrenaline or enoximone.

Table 3

Interaction of Mg2+ concentration and operative priority.

Regression term	Odds ratio	95% CI	p
High versus low Mg2+ concentration	1.31	0.86–1.99	0.21
Urgent versus elective operative priority	2.34	1.42–3.86	0.001
Interaction Mg2+ concentration and operative priority	0.48	0.24–0.99	0.05

There were no statistically significant differences between groups for other clinical outcomes, summarised in Table 4. Geometric means for times to plegia arrest (available for 307 participants in each group) were 51.3 and 51.9 s for the low and high groups, respectively (high group 1% longer, 95% CI: −6% to 9%, p = 0.79).

Table 4

Clinical outcomes and complications for the 5 mmol l−1 and 16 mmol l−1 Mg2+ groups.

	Low group (n = 342)		High group (n = 349)		High versus low groups		pa
	Frequency	%	Frequency	%	Risk ratio	95% CI
Post-CPB vasodilators	5	1.61%	10	3.2%	1.96	0.68–5.66	0.21
Postoperative vasodilators (any)	40	13.2%	36	11.4%	0.87	0.57–1.32	0.50
Postoperative sepsis	11	3.3%	4	1.2%	0.35	0.11–1.10	0.07
Postoperative chest infection	35	10.6%	29	8.5%	0.81	0.51–1.29	0.37
Postoperative wound infection	8	2.4%	11	3.2%	1.34	0.55–3.29	0.53
Postoperative pneumothorax	8	2.6%	10	3.1%	1.20	0.48–3.00	0.70
Neurological complication (any)	3	0.9%	4	1.2%	1.31	0.29–5.79	1.00
Postoperative blood loss (ml)a	600	425–800	600	450–875	1.05b	0.96–1.16	0.26
Red cell transfusion (any)	82	25.0%	80	24.2%	0.97	0.74–1.27	0.82
ICU stay (days)a	1.0	0.91–1.05	1.0	0.94–1.11	1.06b	0.97–1.15	0.19
Postoperative stay (days)a	6	5–8	6	5–8	0.98b	0.93–1.04	0.50
30-day mortality	0	0.00%	5	1.4%	na	na	0.06

CPB: cardiopulmonary bypass; ICU: cardiac intensive care unit; and na: risk ratio not calculable because the outcome did not occur in one or other group. Some outcomes for some patients were missing.

Median and interquartile range.

Ratio of geometric means.

Differences between groups in the frequencies of two clinical outcomes, 30-day mortality and postoperative sepsis, approached statistical significance (p = 0.06 and 0.07, respectively) but were not in the same direction (Table 4). Five participants in Bristol died within 30 days (5/516, 1.0%), four of them in hospital; all deaths were in the high-Mg2+ group (overall 30-day mortality rate, 0.7%; rate in the high group, 1.4%; Fisher's exact test p = 0.06).

From reviewing the medical records, we found no features linking the deaths and no indication that they were attributable to a high-Mg2+ concentration. One patient developed POAF, was cardioverted with amiodarone, discharged in sinus rhythm and then found dead at home. A second patient was in heart failure before the operation and had poor quality target arteries; she had an intra-aortic balloon pump inserted during surgery, came off bypass with much inotropic support, developed biventricular failure and died despite increasing inotropic support and continuous haemofiltration. A third patient developed complete heart block on the 7th postoperative day and was treated by inserting a transvenous pacing wire; he subsequently developed renal failure, sepsis, bowel ischaemia and died from multi-organ failure. A fourth patient had an urgent operation; he had small diffusely diseased coronaries, suffered a perioperative myocardial infarction and subsequently died. The fifth patient had a respiratory arrest.

We used routine data for all of the 516 participants in Bristol to compare longer-term survival (post hoc analysis prompted by the 30-day mortality findings). There were 34 deaths in total up to 5 years after surgery (mean duration of follow-up 3.0 years, 18 and 16 deaths in the low and high groups, respectively, log-rank test, p = 0.64). The overall proportion dying by 3 years was 5.7% (95% CI 4.0–8.3%).

Blood electrolytes and biochemical markers

Varying numbers of patients had data for these outcomes (n = 72 for troponin to n = 414 for Mg2+) because the required blood samples were not always taken or were not available from laboratory databases. Geometric mean concentrations in the two groups are described in Table 5. For Mg2+, there was a highly significant interaction of group and time after surgery (Fig. 2(A)). Mg2+ concentration was significantly higher in the high than in the low group at 4 h and at 12 h. Thereafter, there were no differences between groups.

Table 5

Geometric means (and standard errors) for 5 mmol l−1 and 16 mmol l−1 Mg2+ groups, ratios of means and confidence intervals.

Variable	Low groupa	High groupa	Ratio	95% CI	p
Magnesium (mmol l−1)	(n = 199)	(n = 215)			<0.001b
Pre-op	0.87 (1.01)	0.85 (1.01)
4 h	0.84 (1.01)	1.08 (1.01)	1.29	(1.25–1.32)	<0.001
12 h	0.89 (1.01)	0.93 (1.01)	1.04	(1.01–1.08)	0.01
24 h	0.90 (1.01)	0.88 (1.01)	0.98	(0.95–1.01)	0.18
48 h	0.87 (1.01)	0.87 (1.01)	0.99	(0.97–1.02)	0.58
Troponin I (ng ml−1)	(n = 40)	(n = 32)			0.23b
Pre-op	0.02 (1.15)	0.02 (1.11)			na
4 h	0.27 (1.11)	0.26 (1.12)	0.96		na
12 h	0.31 (1.12)	0.23 (1.11)	0.73		na
24 h	0.23 (1.15)	0.15 (1.14)	0.63		na
48 h	0.14 (1.16)	0.08 (1.14)	0.61		na
Estimate of overall effect			0.72	(0.55–0.94)	0.02
Potassium (mmol l−1)	(n = 209)	(n = 218)			0.38b
Pre-op	4.28 (1.01)	4.28 (1.01)			na
4 h	4.26 (1.01)	4.23 (1.01)	0.99		na
12 h	4.22 (1.00)	4.24 (1.00)	1.00		na
24 h	4.24 (1.00)	4.23 (1.01)	1.00		na
48 h	4.11 (1.01)	4.16 (1.01)	1.01		na
Estimate of overall effect			1.00	(0.99–1.01)	0.67
Lactate (mmol l−1)	(n = 148)	(n = 168)			0.76b
Pre-op	1.04 (1.04)	1.04 (1.03)			na
4 h	1.34 (1.03)	1.41 (1.02)	1.05		na
12 h	1.16 (1.04)	1.17 (1.04)	1.01		na
24 h	1.03 (1.04)	1.10 (1.04)	1.06		na
48 h	0.97 (1.07)	0.97 (1.07)	1.00		na
Estimate of overall effect			1.03	(0.95–1.11)	0.44
Creatinine (μmol l−1)	(n = 207)	(n = 218)			0.20b
Pre-op	103.0 (1.02)	104.5 (1.01)			na
4 h	94.3 (1.01)	93.4 (1.01)	0.99		na
12 h	103.6 (1.01)	102.8 (1.01)	0.99		na
24 h	108.8 (1.02)	108.4 (1.02)	1.00		na
48 h	105.2 (1.02)	102.2 (1.02)	0.97		na
Estimate of overall effect			0.99	(0.96–1.02)	0.38

Means and standard errors have been back-transformed into the units of measurement. Serum concentrations after surgery were adjusted for the preoperative concentrations.

p value for the test of interaction. Separate ratio estimates are only reported when this value was <0.05; otherwise, they are not applicable (na).

Fig. 2

Serum concentrations of (A) magnesium (mmol l−1) and (B) troponin I (ng ml−1) during the first 48 h postoperatively.

There was no interaction of group and time after surgery for any of the other biomarkers. The overall effect estimate for high versus low concentration was significant only for troponin I (Fig. 2(B)). The ratio of geometric means for troponin I was 0.72 (95% CI 0.55–0.94, p = 0.02), representing a 28% relative reduction in the high group.

Discussion

Main findings

This trial found no difference in the frequency of POAF between high and low concentrations of Mg2+, or any differences between other clinical outcomes. Troponin release was reduced by almost one-third in the high compared with the low group. The overall frequency of POAF was slightly higher than previously observed (32% vs 26%), despite a lower overall percentage of urgent patients (32% vs 45%) [7].

Strengths and limitations

The findings of the trial are reported in accordance with guidelines [14]. The trial was about 10 times larger than all but one previous trial of Mg2+ supplementation of cardioplegia. Random allocation was concealed and only the perfusionist was unblinded, avoiding common biases (www.cochrane-handbook.org, accessed 9 April 2010) [15]. Primary and secondary outcomes were defined in the trial protocol.

The trial had some limitations. The trial recruited over a much longer period than expected, partly because of continuing uptake of off-pump CABG at the coordinating site (≈70% of all isolated CABG in 2007). Funding for the trial was exhausted before completion, limiting the availability of research staff to recruit and follow patients. These factors may limit the applicability of the trial findings, although the characteristics of the participants do not suggest marked selection.

Outcomes, especially biomarkers, were missing for several patients but to the same extent in both groups, supporting the view that data were lost because research staffs were often not available. The original intention was to collect these markers in a consecutive subsample of trial participants, but this was not feasible. Data missing because staffs were not available would have caused a loss of power, but would have been very unlikely to introduce bias.

The first 40 randomised participants were excluded. This decision did not introduce bias because it was made before any analyses were carried out and because the number of excluded patients was chosen to coincide with the end of a randomisation ‘block’. Ten subsequent participants were also excluded because the level of supplementation administered was not recorded; however, their allocations were balanced, suggesting that these data were missing at random. Twenty-four participants crossed over from one group to the other. As the low concentration was usual practice during the period of the trial, and 17 of these participants received low supplementation instead of the high level allocated, we believe that crossovers represent genuine errors. Crossovers would have biased the findings towards the null hypothesis.

Findings in the context of other literature

POAF is commonly regarded as a benign, self-limiting complication. However, it has been associated with longer intensive care unit (ICU) stay, an increased risk of ICU re-admission, prolonged hospital stay and increased hospital costs [16]. Atrial fibrillation is also the leading cause of hospital re-admission after cardiac surgery [17]. Because of its clinical and economic impact, interventions to reduce the risk of POAF are important.

Four small trials (sample sizes from 18 to 70) have compared cold-blood cardioplegia (CBC) either supplemented with Mg2+ or not with concentrations varying from 3 to 18 mmol l−1 [18-21]. Three trials compared the risk of POAF, with frequencies of POAF in control groups of 30–40% [18,20,21]. A random effects meta-analysis of these trials suggests that Mg2+ supplementation of CBC reduces the odds of POAF by 62% (pooled OR = 0.38, 95% CI: 0.17–0.85, p = 0.02). Publication bias might explain this finding but is difficult to assess with only three trials. Three trials have compared conventional IAWBC with IAWBC supplemented with Mg2+ (concentrations varying from 2 to 5 mmol l−1) [6,7,22], but only one trial reported frequencies of POAF of 22% and 29%, respectively, in Mg2+ and control groups (p = 0.07) [7].

We found an interaction between Mg2+ concentration and operative priority, with the high concentration (16 mmol l−1) reducing the odds of POAF in urgent cases. Our previous trial also suggested that Mg2+ supplementation reduced POAF in urgent but not in elective cases (interaction test, p = 0.10) [7]. However, these findings are in fact contradictory, as the low group in the present trial had the same Mg2+ concentration (5 mmol l−1) observed to be beneficial in the previous trial. Consequently, we suspect that both interactions are spurious.

Frequencies of POAF in both groups in this trial (32–33%) were more similar to the POAF frequency for the non-supplemented (29%) than the supplemented (5 mmol l−1; 22%) group reported previously [7]. The increasing overall frequency of POAF over time (32% in this trial compared with 26% previously) is most likely to be attributable to the increasing age of participants having CABG procedures [23].

With respect to myocardial protection against ischaemia and reperfusion, two trials using CBC measured troponin I [19,21] and two measured creatine kinase [19,20]; one trial using IAWBC measured troponin I and creatine kinase [22]. All four trials observed less release of cardiac enzymes in groups with moderate Mg2+ supplementation [19-22] but not at a high concentration of 16–18 mmol l−1 [19]. This evidence points to a consistent protective effect on the myocardium, although the optimal level of supplementation is unclear.

With the exception of 30-day mortality, the overall frequencies of postoperative complications and the difference in frequencies between groups (Table 4) did not cause concern. However, the difference in the risk of dying within 30 days initially caused alarm. Deaths in hospital were carefully reviewed without identifying any common features or clinical signs or symptoms that were attributed to high Mg2+ (see the section ‘Results’). Mg2+ concentrations similar to that used in this trial have been studied previously in CBC without any report of harm [18,19].

Based on the literature as a whole, we are no longer of the opinion that cardioplegia for IAWBC should be supplemented with Mg2+ ‘to reduce the risk of POAF’. Neither of the two large trials done by our research group found a clear benefit of Mg2+ supplementation of IAWBC. There remains the possibility that Mg2+ supplementation of CBC is important to reduce POAF, given the findings of the three small trials, although these three trials were very small (160 patients in total). However, our data support the view that IAWBC should be supplemented with Mg2+ ‘to reduce myocardial injury’, although we accept that this effect is not yet proven and that the optimal dose needs to be established.

Conclusions

High level of Mg2+ supplementation of IAWBC is safe and may reduce cardiac injury, but does not reduce the frequency of POAF. Future research should focus on the role of Mg2+ in cardioplegia in reducing cardiac injury and establish the optimal level of supplementation.