Beneficial effects of levosimendan on survival in patients undergoing extracorporeal membrane oxygenation after cardiovascular surgery

Abstract

Background

The impact of levosimendan treatment on clinical outcome in patients undergoing extracorporeal membrane oxygenation (ECMO) support after cardiovascular surgery is unknown. We hypothesized that the beneficial effects of levosimendan might improve survival when adequate end-organ perfusion is ensured by concomitant ECMO therapy. We therefore studied the impact of levosimendan treatment on survival and failure of ECMO weaning in patients after cardiovascular surgery.

Methods

We enrolled a total of 240 patients undergoing veno-arterial ECMO therapy after cardiovascular surgery at a university-affiliated tertiary care centre into our observational single-centre registry.

Results

During a median follow-up period of 37 months (interquartile range 19–67 months), 65% of patients died. Seventy-five per cent of patients received levosimendan treatment within the first 24 h after initiation of ECMO therapy. Cox regression analysis showed an association between levosimendan treatment and successful ECMO weaning [adjusted hazard ratio (HR) 0.41; 95% confience interval (CI) 0.22–0.80; P=0.008], 30 day mortality (adjusted HR 0.52; 95% CI 0.30–0.89; P=0.016), and long-term mortality (adjusted HR 0.64; 95% CI 0.42–0.98; P=0.04).

Conclusions

These data suggest an association between levosimendan treatment and improved short- and long-term survival in patients undergoing ECMO support after cardiovascular surgery.

Editor's key points.

• The effects of levosimendan on outcomes in patients with cardiogenic shock after cardiac surgery are not known.

• In this retrospective observational database analysis, patients requiring ECMO after cardiac or aortic procedures were studied.

• Levosimendan therapy was associated with an increased incidence of weaning from ECMO and reduced short- and long-term mortality.

• However, there were a variety of surgical procedures and indications for ECMO.

• Prospective randomized controlled trials are needed before firm conclusions can be made.

Transient myocardial stunning after cardiovascular surgery is associated with low cardiac output, demanding pharmaceutical or mechanical support during and after weaning from cardiopulmonary bypass.^1,2 This phenomenon is a potentially reversible, but is a life-threatening complication in this vulnerable patient population.^3,4 An imbalance of cardiomyocyte calcium homeostasis is critically involved in myocardial stunning, characterized by a decreased responsiveness of the contractile proteins to calcium and an excitation–contraction uncoupling defect.⁵ Consequently, maintenance of calcium homeostasis might represent a promising therapeutic target in cardiovascular surgical patients with low cardiac output. The calcium-sensitizing inotropic agent levosimendan,⁶ which has been approved for treatment of acutely decompensated heart failure, might therefore be a potential therapeutic option to improve myocardial function in these patients. Indeed, the administration of levosimendan has been shown to reduce the incidence of heart failure and facilitates weaning from cardiopulmonary bypass.^7–10 However, data about the effect of levosimendan treatment on survival are inconclusive.¹¹ A potential mechanism compromising the beneficial effects of levosimendan on clinical outcome in this critical patient population might be the vasodilatory properties¹² that are responsible for arterial hypotension and an increased requirement for vasopressor agents in patients treated with levosimendan.^9,13 Furthermore, a higher incidence of supraventricular and ventricular tachycardia has been observed during levosimendan administration.^13,14 Given that salvage of organ perfusion and tissue oxygenation is of utmost importance in the sensitive postoperative phase, the deleterious effects might have affected clinical outcomes.

In patients with postcardiotomy cardiogenic shock, refractory to conventional medical therapies, veno-arterial extracorporeal membrane oxygenation (ECMO) support is frequently used to restore and maintain adequate end-organ perfusion.^15,16 It is tempting to speculate that the positive inotropic effects of levosimendan are more efficiently translated into improved survival when adequate end-organ perfusion is ensured by concomitant ECMO therapy. However, the effect of levosimendan treatment on survival in patients undergoing ECMO support has not yet been investigated. We therefore analysed the impact of levosimendan treatment on survival and failure of ECMO weaning post hoc in our registry of adult patients undergoing veno-arterial ECMO support after cardiovascular surgery.

Methods

Study population

We enrolled patients undergoing veno-arterial ECMO support after cardiovascular surgery between September 2003 and June 2014 into our observational single-centre registry. The primary inclusion criteria were initiation of veno-arterial ECMO support after cardiovascular surgery. The exclusion criterion was age <18 yr. All patients were recruited at the Vienna General Hospital, a university-affiliated tertiary center as previously described.¹⁶ The study protocol was approved by the Ethics Committee of the Medical University of Vienna and is in accordance with the Declaration of Helsinki.

Levosimendan administration and ECMO management

Levosimendan was administered according to the individual clinical judgement of the on-shift intensive care physician within the first 24 h after initiation of ECMO therapy. All patients in the levosimendan group received levosimendan 12.5 mg in 50 ml of NaCl 0.9% given as a continuous infusion without an initial levosimendan bolus according to the local standard protocol. The ECMO therapy was initiated in patients with clinical signs of severe cardiogenic shock, such as systolic arterial hypotension (<80 mm Hg), and signs of end-organ failure, anaerobic metabolism, and metabolic acidosis despite optimized supportive measures (i.e. inotropes, fluids, and intra-aortic balloon pump). The ECMO system consisted of a centrifugal pump console (Bio-Console560; Medtronic, USA or Cardiohelp system; Maquet, Rastatt, Germany) with a membrane oxygenator (Affinity-NTTM; Medtronic, Minneapolis, MN, USA or HLS module advanced; Maquet, Rastatt, Germany). All components of the extracorporeal oxygenation system were coated with heparin. The ECMO circuit was changed if clots or significant fibrin deposits were present or if blood oxygenation declined drastically. Mechanical ventilation was reduced during ECMO support, with peak airway pressures <25 cm H₂O, respiratory tidal volumes between 6 and 8 ml kg⁻¹, and fractional inspired oxygen on the oxygenator regulated to maintain a target oxygen partial pressure of 80–100 mm Hg. No additional left-sided venting or atrial septostomy was performed. The ECMO device was routinely checked on a 24 h basis by an experienced perfusionist and the on-shift intensive care physician.

Clinical definitions and study end points

Cardiovascular risk factors were recorded according to the respective guidelines. The glomerular filtration rate was calculated using the MDRD formula. Left ventricular function was categorized into normal [left ventricular ejection fraction (LVEF) ≥55%], mildly reduced (LVEF 45–54%), moderately reduced (LVEF 30–44%), and severely reduced (LVEF <30%). Furthermore, the simplified acute physiology score 3 (SAPS-3), the sequential organ failure assessment (SOFA) score, and European system of cardiac operative risk evaluation (EuroSCORE) were computed.^17,18 Thirty-day mortality was defined as the primary study end point and obtained by screening of the national register of death. Secondary end points were long-term mortality and failure of ECMO weaning. Failure of ECMO weaning was defined as death during ECMO support or death within 24 h after ECMO removal.

Statistical methods

Discrete data are presented as the count and percentage and analysed by using a χ² test. Continuous data are given as the median and interquartile range (IQR) and compared using Mann–Whitney statistics. The effect of levosimendan on outcome was assessed post hoc. Cox proportional hazard regression analysis was applied to assess the effect of levosimendan treatment on survival and failure of ECMO weaning. Results are expressed as the hazard ratio (HR) with the respective 95% confidence interval (CI). To account for potential confounding effects, we calculated the risk for death adjusted for for age, sex, SAPS-3, SOFA score, hypertension, diabetes, maximal norepinephrine dose within the first 24 h, left ventricular function, duration of ECMO support, and type of cardiovascular surgery.

Interactions between levosimendan therapy and all variables included in the multivariable model were excluded by entering interaction terms in the Cox proportional hazard regression models. We additionally tested for collinearity in the multivariate model using the variance inflation factor. Two-sided P-values <0.05 were taken to indicate statistical significance. The STATA11 software package (StataCorp, College Station, TX, USA) and SPSS 23.0 (IBM, Armonk, NY, USA) were used for all analyses.

Results

Baseline characteristics

We included 240 patients, with a median age of 65 yr (IQR 55–72 yr), undergoing ECMO support after cardiovascular surgery between September 2003 and June 2014. Seventy-one per cent of patients (n=171) were male. The median SAPS-3 and median EuroSCORE of the study population were 43 (IQR 36–51) and 10 (IQR 8–13), respectively. The ECMO support was initiated in 59 patients after valve surgery, in 24 after coronary artery bypass graft surgery, in 56 after combined coronary artery bypass graft and valve surgery, in 51 patients after cardiac transplantation, in 21 patients after ventricular assist device implantation, in 17 after aortic reconstruction, and in 12 after other cardiovascular surgeries. Indications for ECMO implantation were weaning failure from cardiopulmonary bypass (60%), postoperative cardiogenic shock (20%), immediate post-transplant cardiac graft failure (6%), postoperative respiratory failure (4%), postoperative bleeding/tamponade with cardiogenic shock (4%), and miscellaneous conditions (6%). The ECMO implantation was performed femoral–femoral in 43% of patients, subclavian–femoral in 47% of patients, and central–femoral in 10% of patients.

In total, 75% of patients received levosimendan during the first 24 h after ECMO implantation. Figure 1 illustrates the number of patients in each treatment arm in each year of the study. In brief, ECMO patients treated with levosimendan had a higher EuroSCORE (P<0.001) and SAPS-3 (P=0.02), had a lower left ventricular ejection fraction (P=0.04), and required higher doses of norepinephrine (P<0.001) and dobutamine (P=0.014). Levosimendan administration was independent from time of study enrolment (P-value for interaction=0.84). A detailed comparison of patients with and without levosimendan is given in Table 1.

Fig 1.

Number of patients on extracorporeal support treated with levosimendan from 2003 to 2014.

Table 1.

Baseline characteristics of total extracorporeal membrane oxygenation study population (n=240) according to study group. ALT, alanine aminotransferase; AST, aspartate aminotransferase; ECMO, extracorporeal membrane oxygenation; EuroSCORE, European system of cardiac operative risk evaluation; $F{\mathrm{I}}_{{\mathrm{O}}_2}$, fractional inspired oxygen; GFR, glomerular filtration rate; ICU, intensive care unit; IQR, interquartile range; γ-GT, gamma-glutamyl transferase; ScVO₂, central venous oxygen saturation. Bold typeface indicates statistical significance (P<0.05)

	Levosimendan (n=179)	Control (n=61)	P-value
Baseline characteristics at hospital admission
Age [median (IQR); yr]	65 (56–72)	65 (52–70)	0.39
Male sex [n (%)]	133 (74)	38 (63)	0.11
EuroSCORE [additive; median (IQR); points]	11 (8–13)	9 (7–10)	<0.001
Duration of procedure [median (IQR); h]	7.8 (6.2–9.5)	8.3 (5.9–10.1)	0.82
Hypertension [n (%)]	129 (72)	40 (66)	0.43
Diabetes [n (%)]	53 (30)	13 (21)	0.23
Hypercholesterolaemia [n (%)]	98 (55)	26 (43)	0.13
Coronary artery disease [n (%)]	93 (52)	31 (51)	0.97
Left ventricular function
Moderately reduced [n (%)]	30 (17)	5 (8)	0.11
Severely reduced [n (%)]	77 (43)	17 (28)	0.04
Creatinine [median (IQR); mg dl−1]	1.5 (1.1–1.8)	1.3 (1.0–1.6)	0.12
Estimated GFR [median (IQR); ml min−1 (1.73 m)−2]	50 (36–68)	56 (43–70)	0.17
Blood urea nitrogen [median (IQR); mg dl−1]	25 (18–41)	23 (19–31)	0.39
Total cholesterol [median (IQR); mg dl−1]	139 (105–179)	145 (98–183)	0.64
Total bilirubin [median (IQR); mg dl−1]	1.1 (0.7–1.6)	0.8 (0.6–1.6)	0.31
ASAT [median (IQR); U l−1]	34 (23–68)	31 (25–63)	0.93
ALAT [median (IQR); U l−1]	26 (17–46)	29 (21–47)	0.25
γ-GT [median (IQR); U l−1]	55 (31–111)	51 (33–105)	0.75
C-reactive protein [median (IQR); mg dl-1]	1.0 (0.3 4.4)	1.0 (0.3–3.5)	0.51
Leucocytes [median (IQR); 1000 μl−1]	7.8 (5.9–10.7)	8.1 (6.8–11.4)	0.12
Platelets [median (IQR); 1000 μl−1]	185 (132–243)	197 (136–236)	0.59
Post-ECMO implantation (first 24 h)
SAPS-3 [n (%)]	45 (36–52)	40 (36–44)	0.02
ECMO flow [median (IQR); l min−1]	3.2 (2.5–4.3)	3.9 (3.0–4.4)	0.21
ECMO rotation [median (IQR); rpm]	3000 (2450–3590)	3100 (2670–3460)	0.93
ECMO gas flow [median (IQR); l min−1]	2.5 (2.0–3.0)	3.0 (2.0–4.0)	0.32
ECMO $F{\mathrm{I}}_{{\mathrm{O}}_2}$ [median (IQR); %]	65 (60–90)	70 (60–100)	0.67
ECMO duration [median (IQR); days]	4 (3–7)	4 (2–7)	0.33
Haemodynamic parameters (at ICU admission)
Mean arterial pressure [median (IQR); mm Hg]	72 (65–79)	73 (65–80)	0.54
Cardiac output [median (IQR); l min−1]	3.7 (2.8–5.1)	3.8 (3.0–7.0)	0.48
$S\text{c}{\text{V}}_{{\mathrm{O}}_2}$ [median (IQR); %]	71 (65–77)	69 (61–75)	0.31
Central venous pressure [median (IQR); mm Hg]	14 (11–16)	14 (12–17)	0.73
Medication (first 24 h post-ECMO)
Norepinephrine [n (%)]	177 (99)	58 (95)	0.07
Norepinephrine (maximal dose; median (IQR); μg kg−1 min−1]	0.33 (0.17–0.60)	0.21 (0.09–0.37)	<0.001
Dobutamine [n (%)]	169 (94)	49 (80)	0.003
Dobutamine (maximal dose; median (IQR); μg kg−1 min−1]	5.00 (3.21–7.18)	4.04 (1.49–6.67)	0.014
Vasopressin [n (%)]	74 (41)	11 (18)	0.001
Vasopressin [median (IQR); U h−1]	3.0 (2.0–4.0)	4.0 (2.3–4.4)	0.17
Dopamine [n (%)]	5 (3)	5 (8)	0.06
24 h fluid balance [median (IQR); ml]	5117 (3909–7100)	4922 (3747–7184)	0.62

Clinical outcomes

During a median follow-up time of 37 months (IQR 19–67 months), 65% of patients (n=156) died. We detected a significant protective effect of levosimendan treatment within the first 24 h after initiation of ECMO therapy on 30 day mortality, with a crude HR of 0.61 (95% CI 0.39–0.96; P=0.03) that was even more pronounced after multivariate adjustment, with an adjusted HR of 0.52 (95% CI 0.30–0.89; P=0.016; Table 2). When assessing the effect of levosimendan on long-term mortality, we detected a crude HR of 0.77 (95% CI 0.54–1.09; P=0.14) that reached statistical significance after multivariate adjustment with an HR of 0.64 (95% CI 0.42–0.987; P=0.04; Table 2). Furthermore, we did not detect a significant collinearity and there was no significant interaction between levosimendan and any variable in the multivariate confounder model (data not shown). Failure of weaning from ECMO support occurred in 23% of patients (n=55). Levosimendan treatment was associated with a reduced risk of ECMO weaning failure, with a crude HR of 0.54 (95% CI 0.31–0.93; P=0.03) despite the lower ejection fraction and higher doses of catecholamines. This effect of levosimendan was even more pronounced after multivariate adjustment, with an adjusted HR of 0.41 (95% CI 0.22–0.80; P=0.008; Table 2). Adjusted Cox survival curves further illustrate the significant decrease of ECMO weaning failure (P=0.008; Supplementary Fig. S1A), 30 day (P=0.016; Supplementary Fig. S1B) and long-term mortality (P=0.04; Fig. 2) in patients receiving levosimendan treatment within the first 24 h after initiation of ECMO therapy. Furthermore, there was no beneficial association between survival and the maximal dose of dobutamine (30 day survival, P=0.83; long-term survival, P=0.41), dopamine (30 day survival, P=0.26; long-term survival, P=0.43), or vasopressin (30 day survival, P=0.10; long-term survival, P=0.46) within the first 24 h. In contrast, we detected an adverse association between the maximal dose of norepinephrine within the first 24 h and survival, with a crude HR per 1 sd increase of 1.33 (95% CI 1.11–1.58; P=0.02) for 30 day mortality and of 1.25 (95% CI 1.07–1.45; P=0.004) for long-term mortality.

Table 2.

Unadjusted and adjusted Cox proportional hazard model analysing the effect of levosimendan treatment on ECMO weaning failure, 30 day mortality, and long-term mortality. Hazard ratios are adjusted for all variables in the clinical confounder model (i.e. for age, sex, SAPS-3, SOFA score, hypertension, diabetes, maximal norepinephrine dose within first 24 h, left ventricular function, duration of ECMO support, and type of cardiovascular surgery). CI, confidence interval; ECMO, extracorporeal membrane oxygenation; HR, hazard ratio; SAPS-3, simplified acute physiology score 3; SOFA, sequential organ failure assessment

	Crude HR (95% CI)	P-value	Adjusted HR (95% CI)	P-value
ECMO weaning failure	0.54 (0.31–0.93)	0.03	0.41 (0.22–0.80)	0.008
30 day mortality	0.61 (0.39–0.96)	0.03	0.52 (0.30–0.89)	0.016
Long-term mortality	0.77 (0.54–1.09)	0.14	0.64 (0.42–0.98)	0.04

Discussion

These data suggest a favourable effect of levosimendan on clinical outcome in patients with veno-arterial ECMO support after cardiovascular surgery. We identified an independent protective effect of levosimendan treatment on short- and long-term mortality. Furthermore, levosimendan-treated patients were more successfully weaned from ECMO despite a more pronounced risk profile reflected by a higher median SAPS-3 and EuroSCORE.

Low cardiac output is frequently observed in patients undergoing cardiovascular surgery.^1,2 Transitory ECMO support represents a valuable therapeutic option to ensure adequate organ perfusion until myocardial recovery. However, the risk of ECMO weaning failure and death is still significant in this vulnerable patient population.^19,20 Considering the pivotal role of a disturbed myocyte calcium homeostasis in myocardial stunning,⁵ levosimendan might be the appropriate therapeutic compound because it increases myocardial contraction by enhancing myofilament responsiveness to calcium, with a neutral effect on oxygen consumption.⁶ Several small studies have suggested a benefit of levosimendan therapy on haemodynamic and clinical parameters in cardiovascular surgery patients.^{7–10,21–24} In detail, Lahtinen and colleagues⁹ demonstrated a reduction of postoperative heart failure in levosimendan-treated patients after high-risk cardiac surgery. The Randomized EValuation of Intravenous LeVosimendan Efficacy (REVIVE) I and II study¹³ confirmed this finding by showing favourable symptomatic effects of levosimendan administration on the short-term clinical course of cardiovascular surgery patients. However, these studies failed to reveal beneficial effects of levosimendan treatment on survival. The risk of death was numerically even higher in levosimendan-treated patients compared with those who were assigned to placebo in the REVIVE II trial.¹³ A recently published meta-analysis concluded that it still remains unclear whether levosimendan treatment has favourable effects on survival.¹¹ Therefore, levosimendan should not be administered generally in cardiovascular surgery patients, but a more individualized treatment decision-making process should be followed, and therefore, specific patient populations that might benefit from levosimendan treatment need to be identified.²⁵ The present study, investigating patients undergoing ECMO support after cardiovascular surgery, indicates for the first time a potentially beneficial effect of levosimendan in this specific setting. We identified an increased short- and long-term survival in levosimendan-treated patients and a higher rate of successful ECMO weaning, despite worse left ventricular function and higher risk scores (EuroSCORE and SAPS-3) compared with patients without levosimendan therapy. Furthermore, levosimendan therapy seems to be safe in patients undergoing ECMO support, because the duration of ECMO therapy was not increased in levosimendan-treated patients.

The mechanisms responsible for cardiac stunning after open heart surgery are not well understood. However, it has been shown that cardioplegic arrest and reperfusion lead to apoptosis of cardiac myocytes.²⁶ In addition to its effects on calcium haemostasis, levosimendan opens mitochondrial ATP-sensitive K⁺ channels (mitoK⁺_ATP channels),^27,28 and cell culture experiments have demonstrated that levosimendan thereby protects rat cardiac myocytes from apoptotic cell death.²⁹ In addition, it has been shown that hypoxic cardiac myocytes produce the inflammatory cytokine interleukin-6 that is increased after reperfusion of the ischaemic heart and that may induce neutrophil-mediated reperfusion injury in the myocardium.³⁰ Levosimendan exerts anti-inflammatory effects and reduces the expression of the inflammatory cytokines interleukin-6 and interleukin-8 in cardiac myocytes in vitro,³¹ and may thereby reduce cardiac injury after open heart surgery.

It is tempting to speculate that protective effects of levosimendan are more efficiently translated into improved survival when end-organ perfusion is ensured by concomitant ECMO therapy. The positive effects of levosimendan on stroke volume, cardiac output, and pulmonary capillary wedge pressure are accompanied by a decrease in mean blood pressure and total peripheral resistance, which may affect end-organ perfusion and, consequently, clinical outcome.³² In line with the haemodynamic properties of levosimendan, an increased requirement for vasopressor agents was observed in levosimendan-treated patients.⁹ High doses of vasopressors are known to increase myocardial oxygen consumption, which in turn can result in myocardial ischaemia, with subsequent injury to hibernating but viable myocardium. The ECMO treatment might alleviate these side-effects because the extracorporeal circuit preserves adequate end-organ perfusion. Additionally, ECMO therapy might compensate for episodes of low cardiac output triggered by arrhythmic events that were more frequently observed in patients treated with levosimendan.¹³ Furthermore, animal studies suggested beneficial effects of levosimendan therapy on pulmonary function by attenuating pulmonary vasoconstriction, hepatic function by reducing hypoxic injury, and renal function by improving glomerular filtration rate.^33–35 Whether these effects mitigate acute organ dysfunction and ultimately translate into improved survival remains unclear, and human data on the protective effects of levosimendan on organ function are eagerly awaited from the ongoing Levosimendan for the Prevention of Acute oRgan Dysfunction in Sepsis (LeoPARDS) trial.³⁶ Finally, the anti-inflammatory properties of levosimendan³¹ may suppress ECMO-related systemic inflammatory responses.³⁷ Therefore, ECMO support and levosimendan therapy might complement each other in their therapeutic actions, which is translated into improved survival.

Considering the confounder-adjusted survival curves (Fig. 2), the beneficial effects of levosimendan did not only persist after successful ECMO weaning but even increased during the follow-up period, suggesting a sustained effect of levosimendan treatment on short-term clinical outcome. This observation might be explained by the long-acting properties of levosimendan mediated by an active circulating metabolite that reaches peak concentrations 5–6 days after discontinuation of the levosimendan infusion in cardiac surgery patients.^8,38 Furthermore, the beneficial effect of levosimendan persisted throughout the entire follow-up period, which can potentially be attributed to the long-term anti-remodelling, anti-apoptotic, antioxidant, and anti-inflammatory cardioprotective properties of levosimendan.^31,39–43

Fig 2.

Confounder-adjusted survival curves of long-term mortality (P=0.04) according to study group.

The primary limitation of our study is its observational design. However, the results of observational, non-randomized studies are needed when randomized clinical trials are not available,⁴⁴ and a randomized clinical trial analysing the impact of levosimendan on survival in patients undergoing ECMO support after cardiovascular surgery is currently not underway and improbable in the near future. Furthermore, given that the decision to administer levosimendan was not made as part of a randomized controlled trial but was left to the discretion of the physician, we cannot exclude a patient selection bias. However, in our clinical routine physicians tend to prescribe levosimendan more frequently in patients with significantly reduced left ventricular function or who present clinically more unstable. In view of the patient baseline characteristics, the beneficial effects of levosimendan might even be underestimated, because levosimendan-treated patients had a significantly higher risk profile, expressed by a higher SAPS-3 and EuroSCORE, compared with patients who did not receive levosimendan therapy. Nevertheless, future randomized clinical trials are warranted to confirm the favourable effect of levosimendan in this specific setting. Finally, there are various dosing regimens for levosimendan, and these results might not apply for other administration regimens.

Conclusion

This observational study suggests an association between levosimendan treatment and improved short- and long-term survival in patients undergoing ECMO support after cardiovascular surgery.

Authors' contributions

Conception of the study: K.D., C.R., C.B., B.S.

Study design: K.D., G.H., I.M.L., G.M., H.K., A.N., M.H., W.S., G.G.

Main coordinator: G.G.

Coordination of the study: G.H., I.M.L., A.N., M.H., W.S.

Data collection: K.D., C.R., L.S., C.B.

Statistical analysis: K.D., B.S., A.N., G.G.

Drafting the manuscript: K.D., C.R., L.S., C.B., A.N., M.H., W.S.

Revision of the manuscript: B.S., G.H., I.M.L., G.M., H.K., G.G.

All authors read and approved the final manuscript.

Supplementary material

Supplementary material is available at British Journal of Anaesthesia online.

Declaration of interest

None declared.