Você está na página 1de 9


Effect of metformin pretreatment on myocardial injury during

coronary artery bypass surgery in patients without diabetes
(MetCAB): a double-blind, randomised controlled trial
Saloua El Messaoudi, Rianne Nederlof, Coert J Zuurbier, Henry A van Swieten, Peter Pickkers, Luc Noyez, Hendrik-Jan Dieker, Marieke J Coenen,
A Rogier T Donders, Annemieke Vos, Gerard A Rongen, Niels P Riksen

Background During coronary artery bypass graft (CABG) surgery, ischaemia and reperfusion damage myocardial Lancet Diabetes Endocrinol 2015
tissue, and increased postoperative plasma troponin concentration is associated with a worse outcome. We investigated Published Online
whether metformin pretreatment limits cardiac injury, assessed by troponin concentrations, during CABG surgery in July 13, 2015
patients without diabetes.
See Online/Comment
Methods We did a placebo-controlled, double-blind, single-centre study in an academic hospital in Nijmegen http://dx.doi.org/10.1016/
(Netherlands) in adult patients without diabetes undergoing an elective on-pump CABG procedure. We randomly S2213-8587(15)00209-0
assigned patients (1:1) in blocks of ten via a computer-generated randomisation sequence to either metformin Departments of Pharmacology-
hydrochloride (500 mg three times per day) or placebo (three times per day) for 3 days before surgery. The last dose Toxicology (S El Messaoudi MD,
was given roughly 3 h before surgery. Patients, investigators, trial staff, and the statistician were all masked to A Vos, Prof G A Rongen MD,
Prof N P Riksen MD), Cardiology
treatment allocation. The primary endpoint was the plasma concentration of high-sensitive troponin I at 6, 12, and (S El Messaoudi, H-J Dieker MD),
24 h postreperfusion after surgery, analysed in the per-protocol population with a mixed-model analysis using all Cardiothoracic Surgery
these timepoints. Secondary endpoints included the occurrence of clinically relevant arrhythmias within 24 hours (Prof H A van Swieten MD,
L Noyez MD), Intensive Care
after reperfusion, the need for inotropic support, time to detubation, duration of stay in the intensive-care unit, and
Medicine (Prof P Pickkers MD),
postoperative use of insulin. This study is registered with ClinicalTrials.gov, number NCT01438723. Human Genetics
(M J Coenen PhD), Department
Findings Between Nov 8, 2011, and Nov 22, 2013, we randomly assigned 111 patients to treatment (57 to metformin for Health Evidence
(A R T Donders PhD), and
and 54 to placebo). Five patients dropped out from the metformin group, and six from the placebo group. 52 patients
Department of Internal
in the metformin group and 48 patients in the placebo group were included in the per-protocol analysis. Geometric Medicine (Prof G A Rongen,
mean high-sensitivity troponin I increased from 0 μg/L to 3·67 μg/L (95% CI 3·06–4·41) with metformin and to Prof N P Riksen), Radboud
3·32 μg/L (2·75–4·01) with placebo at 6 h after reperfusion; 2·84 μg/L (2·37–3·41) and 2·45 μg/L (2·02–2·96), University Medical Center,
Nijmegen, Netherlands; and
respectively, at 12 h; and to 1·77 μg/L (1·47–2·12) and 1·60 μg/L (1·32–1·94) at 24 h. The concentrations did not
Laboratory of Experimental
differ significantly between the groups (difference 12·3% for all timepoints [95% CI –12·4 to 44·1] p=0·35). Intensive Care and
Occurrence of arrhythmias did not differ between groups (three [5·8%] of 52 patients who received metformin vs Anesthesiology, Department of
three [6·3%] of 48 patients who received placebo; p=1·00). There was no difference between groups in the need for Anesthesiology, Academic
Medical Center, Amsterdam,
inotropic support, time to detubation, duration of stay in the intensive-care unit, or postoperative use of insulin. No
Netherlands (R Nederlof MSc,
patients died within 30 days after surgery. Occurrence of gastrointestinal discomfort (mostly diarrhoea) was C J Zuurbier PhD)
significantly higher with metformin than with placebo (11 [21·2%] of 52 vs two [4·2%] of 48 patients; p=0·01). Correspondence to:
Prof Niels P Riksen, Department
Interpretation Short-term metformin pretreatment, although safe, does not seem to be an effective strategy to reduce of Internal Medicine, Radboud
University Medical Center,
periprocedural myocardial injury in patients without diabetes undergoing CABG surgery.
6500 HB Nijmegen, Netherlands
Funding Netherlands Organisation for Health Research and Development and Netherlands Heart Foundation.

Introduction CABG surgery might be improved by strategies that

Coronary artery disease is the leading cause of morbidity reduce myocardial ischaemia-reperfusion injury.4
and mortality worldwide. For most patients with Many preclinical studies report that giving the anti-
multiple vessel coronary artery disease, coronary artery hyperglycaemic drug metformin either before ischaemia
bypass graft (CABG) surgery is the preferred strategy for occurs or at the moment of reperfusion potently reduces
revascularisation.1 However, temporary myocardial myocardial ischaemia-reperfusion injury in the setting of
ischaemia during CABG surgery and the subsequent coronary occlusion, both after one dose5 or when given as
reperfusion that occurs causes damage to the cardio- a chronic regimen of several doses.6 Several intracellular
myocytes, a marker of which is increased postoperative signalling cascades contribute to the cardioprotective
plasma concentration of troponin. Increased troponin effect of metformin in animal studies, such as activation
release within the first 24 h after surgery is independently of AMP-activated protein kinase and the prosurvival
associated with increased intermediate-term and long- kinase Akt, and an increased intracellular formation of
term risk of mortality.2,3 Therefore, the prognosis after adenosine.7–9

www.thelancet.com/diabetes-endocrinology Published online July 13, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00121-7 1


Research in context
Evidence before this study Added value of this study
We searched PubMed with the search terms “metformin AND Our study is the first double-blind, randomised controlled trial
(ischaemia OR reperfusion) AND (heart OR myocard* OR investigating whether metformin reduces cardiac injury during
cardiac)” and noted in the literature a consistent infarct CABG surgery in patients without diabetes. Although
size-limiting effect of metformin in murine models of metformin did activate pivotal prosurvival pathways in atrial
myocardial infarction, in animals with and without diabetes. In tissue, it did not reduce plasma high-sensitivity troponin I
human beings, however, the evidence is scarce: in a concentrations after CABG surgery. This finding is in line with
retrospective analysis, metformin-treated patients with the results of other trials in patients without diabetes, showing
diabetes who had a myocardial infarction had lower plasma that metformin does not reduce atherosclerosis or prevent
creatine kinase and troponin concentrations than patients on post-infarction remodelling despite preclinical evidence in
alternative glucose-lowering agents, but this study design is favour of such cardioprotective effects.
prone to confounding. In a randomised prospective pilot study
Implications of all the available evidence
in patients with the metabolic syndrome, a 1 week treatment
Despite preclinical evidence that metformin reduces myocardial
with metformin reduced plasma high-sensitivity troponin I
ischaemia-reperfusion injury, this strategy does not minimise
after elective percutaneous coronary intervention. This study,
such injury during CABG surgery in patients without diabetes.
however, had an open-label design and assessed patients with
Future research should aim to understand why these preclinical
only a minimum amount of cardiac ischaemia, as suggested by
beneficial cardiovascular effects of metformin do not translate
the very low plasma troponin concentrations. By contrast,
to the patient setting.
metformin did not reduce endothelial ischaemia reperfusion
injury in healthy people. The effect of metformin has never
been investigated in the setting of CABG surgery.

Metformin is the most widely used drug in the concomitant valve surgery. After obtaining written
treatment of type 2 diabetes and is associated with informed consent, we screened patients through history
improved cardiovascular prognosis in patients with taking, physical examination, and laboratory analyses.
diabetes compared with other glucose-lowering agents.10,11 Exclusion criteria were off-pump surgery, diabetes, renal
Most guidelines advise to stop metformin before cardiac insufficiency (modification of diet in renal disease
surgery because of the presumed risk of lactic acidosis.12 glomerular filtration rate <60 mL/min), raised con-
The clinical benefit of this policy, however, is still without centrations of liver enzymes (alanine aminotransferase
evidence. A retrospective analysis of patients with more than three times the upper reference limit), or a
diabetes who underwent CABG even reports that the use documented myocardial infarction within 2 weeks of
of metformin on the day before surgery is associated screening. Additionally, patients taking dipyridamole or
with less overall morbidity than the use of non-metformin xanthine derivatives were excluded because these agents
antihyperglycaemic agents, although cardiac morbidity interfere with endogenous adenosine signalling.
and mortality did not differ between treatments.13
In this study (MetCAB), we investigated whether short- Randomisation and masking
term pretreatment with metformin minimises cardiac Patients were randomly assigned (1:1) in blocks of ten to
injury in patients without diabetes undergoing elective pretreatment with either metformin hydrochloride or
CABG surgery. In accompanying exploratory experi- placebo by an independent computer-generated randomi-
ments, we assessed whether treatment with metformin sation sequence (Department of Pharmacy, Radboud
activates key signalling proteins that are involved in University Medical Center, Nijmegen, Netherlands).
cardioprotection. Patients were enrolled by one of the investigators (SEM).
Patients, investigators, trials staff, and the statistician
Methods were all masked to treatment allocation, and masking
Study design and participants was achieved by overencapsulation of metformin tablets
We did a randomised, placebo-controlled, double-blind, under Good Manufacturing Practice conditions at the
single-centre trial at the Radboud University Medical Department of Pharmacy, and the manufacture of
Center (Nijmegen, Netherlands). The study protocol was matching placebo capsules.
in accordance with the Declaration of Helsinki and was
approved by one of the 24 accredited Dutch Medical Procedures
Research Ethics Committees (CMO Regio Arnhem- We assigned patients to either metformin hydrochloride
For the protocol see http://www. Nijmegen). The protocol is available online. (500 mg three times per day) or placebo (three times per
rimls.nl/themes/vascular- We recruited consecutive patients aged at least 18 years day) for 3 days before surgery. In case of postponed
accepted for an elective on-pump CABG with or without surgery, continuation of study drugs was allowed for a

2 www.thelancet.com/diabetes-endocrinology Published online July 13, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00121-7


maximum of 7 days. Because caffeine is an adenosine- concentration after ¹⁰log transformation was
receptor antagonist, all patients had to abstain from 0·53 (SD 0·41), which we used in our sample size
caffeine consumption for at least 24 h before surgery. All calculation. We calculated that a total sample size of
surgical details are reported in the appendix. 80 patients would provide a power of 90% with an α of See Online for appendix
We measured plasma caffeine and metformin con- 0·05 to detect a 50% reduction in postoperative troponin I
centrations immediately before surgery. High-sensitive release (ie, a reduction from 6·76 μg/L to 3·38 μg/L,
troponin I was measured with the Dimension Vista which equals a reduction in the ¹⁰log-transformed
system of Siemens (Newark, USA). The 99th percentile troponin I concentration from 0·53 μg/L to 0·23 μg/L). A
of the reference range is 0·045 μg/L, and the 10% reduction of 50% was expected on the basis of preclinical
coefficient of variation is less than 0·04 μg/L. Additionally, studies on the infarct size-limiting effect of metformin7,9
we did AMPD1 genotyping, because variation in this and the effect of remote ischaemic preconditioning on
gene is associated with resistance against ischaemia- release of troponin I after CABG surgery.14 To enable
reperfusion injury14 (analytical details in appendix). We analysis of the primary endpoint in patients with the CC
estimated insulin resistance immediately before surgery genotype of AMPD1, which is present in 80% of the white
using the homoeostasis model assessment (HOMA-IR) population,17 we added 20% to the calculated 80 patients,
method, with a HOMA-IR threshold value of at least 2·0 and therefore aimed to include 100 assessable patients in
suggesting insulin resistance.15 our study.
We did two exploratory series of laboratory analyses on Plasma high-sensitivity troponin I was not normally
the right atrial appendage, which was harvested before distributed, and results are therefore presented as
the introduction of the extracorporeal circulation in a geometric means with 95% CIs. We used log trans-
subgroup of patients (appendix). First, we measured formation to correct for these skewed data and a mixed-
recovery of contractile function after simulated ischaemia model analysis to assess differences in postoperative
and reperfusion in the atrial trabeculae, a well-established plasma high-sensitivity troponin I to accommodate the
surrogate marker of ischaemia-reperfusion injury.16 repeated measurements by incorporating a participant-
Additionally, we measured phosphorylation of AMP- dependent intercept; the independent variables in this
activated protein kinase and Akt and translocation of model were time after reperfusion and treatment group.
hexokinase II in atrial trabeculae with western blotting. We assessed possible differences between the two
treatment groups over time by incorporating a time-by-
Outcomes treatment interaction. We did a covariate analysis to
Postoperative high-sensitivity plasma troponin I con-
centration was the primary endpoint, measured before
surgery and at 6, 12, and 24 h after release of the aortic 139 assessed for eligibility

clamp (after surgery). Values at all timepoints were used

to calculate the primary endpoint. 28 excluded
Secondary endpoints were the occurrence of clinically 14 did not meet inclusion or exclusion criteria
relevant arrhythmias during the first 24 h after 8 for logistical reasons
5 withdrew informed consent
reperfusion (retrieved from medical charts), the need for 1 enrolled in a different trial
inotropic support, time to detubation, duration of stay in
the intensive-care unit, use of exogenous insulin,
111 randomised
recovery of contractile function of isolated atrial
trabeculae after ex vivo-simulated ischemia-reperfusion,
and expression and activation of AMP-activated protein
kinase, Akt, and hexokinase II in isolated atrial 57 assigned to metformin 54 assigned to placebo
trabeculae. We did not measure renal ischaemia-
reperfusion injury because of logistical reasons.
For safety monitoring, plasma creatinine (to assess 5 drop outs 6 drop outs
2 off pump procedure 1 off pump procedure
renal function), pH, and lactate concentrations were 1 PCI done instead of CABG 1 PCI done instead of CABG
measured 6, 12, and 24 h after surgery. An independent 1 adverse event (GI discomfort) 1 adverse event (GI discomfort)
1 <eight pills taken 1 <eight pills taken
data safety monitoring board did a prespecified safety 1 for logistical reason
analysis after 50% of patients were enrolled. This analysis 1 had endarterectomy of LAD
did not reveal any concerns with respect to safety.
52 completed study and were included 48 completed study and were included
Statistical analysis in per-protocol analysis in per-protocol analysis
We based the sample size calculation on the plasma
troponin I concentrations at 12 h after CABG surgery in Figure 1: Trial profile
880 patients from our own clinic in the period 2006–08 CABG=coronary artery bypass grafting. GI=gastrointestinal. LAD=left anterior descending artery. PCI=percutaneous
(6·76 μg/L [SD 12·96]). The mean plasma troponin I coronary intervention.

www.thelancet.com/diabetes-endocrinology Published online July 13, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00121-7 3


study the effect of propofol anaesthesia on the effect of contraction of the trabeculae, we calculated developed
metformin. A two-sided p value of less than 0·05 was force (difference between the maximum tension during
regarded as significant. Continuous secondary endpoints contraction and the minimum tension during relaxation).
that were normally distributed are presented as mean This variable was averaged for the last 5 min during
(SD) and analysed using an unpaired student’s t test. equilibration and each subsequent 5 min period. Data
Discrete secondary endpoints were analysed using χ² are expressed as percentage of baseline developed force.
analysis or a Fisher’s exact test in case of few events. A paired student’s t test was used to compare in-vitro
Continuous data that were not normally distributed are treatment of trabeculae with control versus metformin
presented as medians (IQR) and were analysed using a during the last 10 min of final reperfusion. To assess the
Mann-Whitney U test. We calculated correlations effect of in-vivo treatment with metformin, an unpaired
between the plasma metformin concentration and the student’s t test was used to compare vehicle-treated
duration of ischaemia and the primary endpoint with a trabeculae of both groups.
Spearman rank order correlation. Because our study is a All western blot data were normalised to placebo
proof-of-concept study, we did a per-protocol analysis (we (dividing all individual values by the mean of the placebo
defined the per-protocol population as all patients who group) and presented as mean (SE). We used an unpaired
completed the protocol; ie, all patients minus the drop- student’s t test to compare the metformin and placebo
outs). Additionally, we did an intention-to-treat analysis groups. SAS version 9.2 was used for all statistical
of the primary endpoint. analyses. This study is registered with ClinicalTrials.gov,
For the laboratory experiments on the effect of number NCT01438723.
simulated ischaemia-reperfusion on contraction of atrial
trabeculae, all data are presented as mean (SD). For each
Metformin Placebo
(n=52) (n=48)
Metformin Placebo
(n=52) (n=48) (Continued from previous coloumn)
Age 65·2 (8·8) 64·5 (9·8) Cardiovascular risk factors*
Sex Hypertension 30 (58%) 30 (63%)
Male participants 43 (83%) 38 (79%) Smoking 15 (29%) 14 (29%)
Female participants 9 (17%) 10 (21%) Dyslipidaemia 19 (37%) 15 (31%)
BMI (kg/m2) 27·5 (4·1) 27·7 (3·5) Family history of cardiovascular disease 25 (48%) 28 (58%)
(first degree relative, aged <60 years)
Blood pressure (mm Hg)
Systolic 147 (23) 147 (22)
Aspirin 45 (87%) 40 (83%)
Diastolic 82 (11) 83 (11)
β blocker 43 (83%) 40 (83%)
Heart rate (beats/min) 65 (9) 64 (11)
Statin 40 (77%) 40 (83%)
Heart rhythm at screening
ACE inhibitor 21 (40%) 17 (35%)
Sinus rhythm 49 (94·2) 47 (97·9)
Angiotensin-II-receptor antagonist 8 (15%) 10 (21%)
Atrial fibrillation 3 (5·8) 1 (2·1)
Calcium channel blocker 10 (19%) 9 (19%)
Concomitant valve surgery 10 (19%) 11 (22%)
Nitrates 34 (65%) 29 (60%)
Aortic valve 9 (17%) 9 (19%)
Diuretic (loop and thiazide) 15 (29%) 15 (31%)
Mitral valve 1 (2%) 1 (2%)
Clopidogrel 14 (27%) 6 (13%)
Combination aortic and mitral 0 (0%) 1 (2%)
Laboratory values at screening
Cardiovascular history
Glucose (mmol/L) 6·2 (1·7) 6·4 (1·6)
Angina pectoris CSS class
Total cholesterol (mmol/L) 4·7 (0·99) 4·7 (1·3)
I 13 (25%) 10 (21%)
Creatinine (μmol/L) 81·9 (11·9) 81·9 (12·0)
II 18 (35%) 17 (35%)
GFR ( [MDRD]) 77·3 (9·3) 77·1 (9·7)
III 16 (31%) 16 (33%)
AMPD genotype
IV 5 (10%) 5 (10%)
CC 34 (65%) 33 (69%)
Previous myocardial infarction 20 (38%) 12 (25%)
CT 18 (35%) 13 (27%)
Previous CABG 3 (6%) 0
TT 0 (0%) 2 (4%)
Previous PCI 13 (25%) 13 (27%)
Number of investigational pills taken 9 (9–10) 9 (9–10)
Left ventricular ejection fraction at screening
>55% 21 (40%) 17 (35%) Data are number (%), mean (SD), or median (IQR). CSS=Canadian Cardiovascular
35–55% 9 (17%) 11 (23%) Association. CABG=coronary artery bypass grafting. PCI=percutaneous coronary
intervention. ACE=angiotensin-converting enzyme. GFR=glomerular filtration rate.
<35% 1 (2%) 2 (4%)
MDRD=modification of diet in renal disease formula. *According to the treating
Unknown 21 (40%) 18 (38%) physician as noted in the medical files.
(Table 1 continues in next column)
Table 1: Baseline characteristics of the per-protocol population

4 www.thelancet.com/diabetes-endocrinology Published online July 13, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00121-7


Role of the funding source 6 Metformin (n=52)

The funder of the study had no role in study design, data Placebo (n=48)
collection, data analysis, data interpretation, or writing of Difference: 12·3% (95% CI –12·4 to 44·1)
p=0·35 (analysis based on all timepoints)
the report. The corresponding author had full access to 5
all the data in the study and had final responsibility for
the decision to submit for publication

High-sensitivity troponin I (μg/L)


Between Nov 8, 2011, and Nov 22, 2013, 139 consecutive
patients provided informed consent. 28 patients were
ineligible (appendix), therefore we randomly assigned
111 patients to treatment (57 to metformin and 54 to 2
placebo). Five patients dropped out from the metformin
group, and six from the placebo group. 52 patients in the
metformin group and 48 patients in the placebo group 1
were included in the per-protocol analysis (figure 1).
Baseline characteristics of the per-protocol population
did not differ between the two groups (table 1). Surgical 0 3 6 9 12 15 18 21 24
details are described in the appendix. Time after reperfusion (h)
Immediately before surgery, mean plasma metformin (release of aortic clamp)
concentrations were 8·42 μmol/L (SD 5·98) for the
Figure 2: Postoperative high-sensitivity troponin I concentration in plasma
metformin pretreated group (range 1·16–24·91 μmol/L) Data are geometric means (95% CIs).
and 0·0 μmol/L (0·0) for the placebo group. Caffeine
plasma concentrations were less than 5·2 μmol/L in
91 patients; nine had a concentration of more than Metformin Placebo p value
(n=52) (n=48)
5·2 μmol/L (range 5·2–24·2). Analysis of the primary
endpoint did not change after exclusion of these Days in ICU 1 (1–1) 1 (1–1) 0·78

patients (data not shown). On the day of surgery plasma Patients on inotropic support 40 (77%) 31 (65%) 0·08
fasting glucose concentrations did not differ between Duration of inotropic support (h) 15 (7·25–20·0) 14 (0–21) 0·74
groups (metformin group mean 5·3 mmol/L [SD 0·9], Time to detubation (h) 12 (10·0–17·8) 12 (11·0–14·8) 0·71
placebo group 5·2 mmol/L [2·2], p=0·35), nor did Use of insulin after surgery 44 (85%) 41 (85%) 0·94
preoperative plasma fasting insulin concentrations Patients with clinically relevant arrhythmias within 24 h of surgery
(metformin 106·26 pmol/L [75·7], placebo 114·6 pmol/L None 49 (94%) 45 (94%) 1·00
[86·12], p=0·59). HOMA-IR was also similar in both AF/SVT de novo 2 (4%) 0 0·50
groups (metformin 4·8 [SD 8·6], placebo 4·5 [4·6], Sinus bradycardia with need for pacemaker 1 (2%) 0 1·00
p=0·24) and showed insulin resistance in our study AV conduction disturbance 0 0 ··
population. VT or VF 0 2 (4%) 0·23
None of the patients died within the first 30 days after Other 0 1 (2%) 0·48
surgery. Baseline concentrations of high-sensitivity
Data are median (IQR) or n (%). ICU=intensive-care unit. AF=atrial fibrillation. SVT=supraventricular tachycardia.
troponin I were below the detection limit of 0·02 μg/L in AV=atrioventricular. VT=ventricular tachycardia. VF=ventricular fibrillation.
79 patients. In 21 patients baseline high-sensitivity
troponin I concentrations were more than 0·02 μg/L Table 2: Secondary endpoints

(11 [21·2%] of 52 patients in the metformin group and

ten [20·8%] of 48 patients in the placebo group; p=0·97, did not differ between the groups at 6, 12, and 24 h after
Pearson’s chi square test), with a range of 0·21–1·88 μg/L reperfusion (difference 12·3% [95% CI –12·4 to
in the metformin group and 0·03–0·49 μg/L in the 44·1], p=0·35; figure 2). An intention-to-treat analysis
placebo group (p=0·51, Mann-Whitney U test). showed similar results: difference 15·2% (–11·1 to 41·5,
Geometric mean high-sensitivity troponin I increased p=0·26). In the metformin group, metformin con-
from 0 μg/L to 3·67 μg/L (95% CI 3·06–4·41) with centrations did not correlate with high-sensitivity
metformin and to 3·32 μg/L (2·75–4·01) with placebo at troponin I concentrations (r=–0·14, p=0·33). With regard
6 h after reperfusion; to 2·84 μg/L (2·37–3·41) with to the use of propofol anaesthesia (15 [29%] of 52 patients
metformin and 2·45 μg/L (2·02–2·96) with placebo at in the metformin group vs 19 [40%] of 48 in the control
12 h; and to 1·77 μg/L (1·47–2·12) with metformin and group), propofol treatment did not modulate the effect of
1·60 μg/L (1·32–1·94) with placebo at 24 h. We noted no metformin on high-sensitivity troponin I concentration:
significant interaction between time and treatment in a covariate analysis, propofol had no effect (p=0·47),
group (p=0·73), so the interaction was removed from the and we did not find a significant interaction with the
model. The concentrations of high-sensitivity troponin I treatment (p=0·98).

www.thelancet.com/diabetes-endocrinology Published online July 13, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00121-7 5


The AMPD1 34C→T CT or TT genotype was present in

A Patients who received placebo in vivo (n=6)
140 Control (ex vivo) 18 (35%) of 52 patients in the metformin group and in
Metformin (ex vivo) 15 (31%) of 48 patients in the placebo group. No
120 significant interaction between the genotype and the
effect of metformin on high-sensitivity troponin I
concentrations was noted (data not shown).
Contractile force (% of baseline)

The duration of ischaemia (mean 68 min [SD 23; range
24–126]), was correlated with high-sensitivity troponin I
concentrations (r=0·41, p<0·0001).
The occurrence of arrhythmias within 24 h after
surgery, including atrial fibrillation or atrial tachycardia,
did not differ between groups (three [5·8%] of 52 patients
who received metformin vs three [6·3%] of 48 patients
who received placebo; p=1·00). Metformin did not affect
time to detubation, length of stay in an intensive-care
unit, or need for inotropic support. There was no
0 difference between groups in the number of patients
Simulated ischaemia Simulated reperfusion who received insulin treatment after surgery (median
18·9 units [IQR 7·2–31·8] in 85% of patients in the
B Patients who received metformin in vivo (n=10) metformin group vs median 19·7 units [11·1–29·0]) in
140 Control (ex vivo)
Metformin (ex vivo) 85% of patients in the control group; table 2).
p=0·18 No differences were noted in postoperative changes in
kidney function or pH levels between groups. Median
creatinine concentrations at 6, 12, and 24 h after surgery
Contractile force (% of baseline)

were 86 μmol/L (IQR 74–97), 87 μmol/L (74–96), and
79 μmol/L (69–88), respectively, in the metformin group,
80 versus 83 μmol/L (74–99), 79 μmol/L (68–94), and
73 μmol/L (61–91), respectively, in the placebo group
60 (p=0·22). In the metformin group, median lactate
concentrations at 6, 12, and 24 h after surgery were
40 1·4 mmol/L (IQR 1·2–2·0), 1·8 mmol/L (1·3–2·6), and
1·8 mmol/L (1·2–2·3), respectively, versus 1·2 mmol/L
20 (1·0–2·2), 2·0 mmol/L (1·5–2·7), and 1·9 mmol/L
(1·4–2·7) in the placebo group (p=0·21). Bleeding
0 complications or need for additional surgical intervention
Simulated ischaemia Simulated reperfusion did not differ between the groups. Common adverse
events included the occurrence of gastrointestinal
C Trabeculae treated with control ex vivo discomfort (mostly mild and self-limiting diarrhoea),
140 Placebo pretreatment (in vivo)
Metformin pretreatment (in vivo)
which was higher in the metformin group than in the
p=0·38 placebo group (11 [21·2%] of 52 vs two [4·2%] of
120 48 patients; see appendix for other details about other
adverse events).
Contractile force (% of baseline)

100 Post-ischaemic recovery of contractile function was

determined in atrial trabeculae of 16 patients (ten
80 receiving metformin, six receiving placebo). The baseline
characteristics of the patients in whom these studies
60 were undertaken are summarised in the appendix.
Baseline contractile force was similar for all groups.
40 Simulated ischaemia reduced contractile force to a

20 Figure 3: The course of contractile force over time

Data are mean (SD). Contractile force from trabeculae harvested during surgery
from six patients randomly assigned to pretreatment with placebo (A) and ten
patients randomly assigned to pretreatment with metformin (B); one trabecula
0 30 60 90 120 150 180 210
from each patient was treated ex vivo with metformin and the other was treated
Time (min)
ex vivo with control (dimethyl sulfoxide). (C) Comparison of trabeculae treated
Simulated ischaemia Simulated reperfusion with vehicle ex vivo from patients pretreated with metformin versus placebo
before surgery.

6 www.thelancet.com/diabetes-endocrinology Published online July 13, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00121-7


similar extent in trabeculae of patients pretreated with however, had an open-label design and only a minimum
metformin or placebo before surgery. Post-ischaemic amount of cardiac ischaemia, as substantiated by the
recovery of contractile function (percentage of baseline very low troponin I concentrations after percutaneous
function) did not differ between the control trabeculae coronary intervention. By contrast, metformin did not
(ie, the trabeculae treated with vehicle in vitro) of the reduce endothelial ischaemia-reperfusion injury in
patients treated before surgery with metformin or healthy people measured with flow-mediated dilation
placebo (30% [SD 14] vs 24% [10], p=0·38; figure 3). after induction of forearm ischaemia by upper-arm cuff
Irrespective of treatment before surgery, additional inflation.20 Our finding that the cardioprotective effects of
metformin added ex vivo did not enhance recovery of metformin in preclinical studies do not translate into a
contractile function after ischaemia-reperfusion reduction in troponin I concentration after CABG are in
(figure 3). Atrial tissue of 22 patients (12 in the placebo line with findings from other randomised controlled
group, ten in the metformin group) was analysed for trials21,22 that showed that metformin does not limit
phosphorylated AMP-activated protein kinase, atherosclerosis or post-infarction remodelling in patients.
phosphorylated Akt, and hexokinase II content by In patients with established coronary heart disease but
western blot (appendix). Compared with control, without diabetes, long-term treatment with metformin
metformin treatment increased phosphorylated AMP- tended to increase the concentration of circulating high-
activated protein kinase by a mean of 10% (SE 3; p=0·03), sensitivity troponin T at 6, 12, and 18 months of treatment
without an effect on total AMP-activated protein kinase. to a mean of 1·0 pg/mL (based on all timepoints), 95% CI
Additionally, compared with control, phosphorylated Akt 0·0–1·9, p=0·05.21
was increased by a mean of 28% (SE 7; p=0·005). We These findings raise the question of why the beneficial
noted no effect of metformin on total Akt, total effects of metformin on cardiac ischaemia-reperfusion
hexokinase II, or mitochondrial-bound hexokinase II injury in animal models do not translate to the clinical
(appendix). setting. First, postoperative troponin I concentration
might not be an accurate marker of myocardial
Discussion ischaemia-reperfusion injury. However, we noted a
In this study, metformin did not limit myocardial significant association between duration of myocardial
ischaemia-reperfusion injury, detected with plasma high- ischaemia and troponin I concentration. Moreover,
sensitivity troponin I concentration, in patients without many previous studies report that interventions that
diabetes undergoing CABG surgery. Additionally, specifically target ischaemia-reperfusion injury reduce
metformin did not affect any of the secondary clinical the concentration of troponin I after CABG surgery.4,14
endpoints. In accordance with these clinical endpoints, Second, the duration or dose, or both of metformin in
metformin did not improve the ex-vivo post-ischaemic clinical trials might be insufficient to activate
recovery of contractile function in isolated atrial prosurvival pathways in cardiac tissue. However, in our
trabeculae, a well established surrogate endpoint for study, ex vivo experiments show that metformin
ischaemia-reperfusion injury. pretreatment did activate pivotal prosurvival proteins.
The search for direct cardioprotective effects of Additionally, a much higher dose of metformin did not
metformin has been fuelled by the findings that patients affect ischaemia-reperfusion injury in isolated atrial
with type 2 diabetes given metformin have a better trabeculae. In murine models of myocardial infarction,
cardiovascular outcome than patients given other substantially lower doses of metformin induced pro-
glucose-lowering agents, despite similar glycaemic found cardioprotection, although plasma concen-
control.10,11 Additionally, many preclinical studies report trations of metformin were not reported in these
beneficial effects of metformin on atherosclerosis studies.7 With regard to the duration, both one dose and
development, cardiac remodelling, and myocardial a chronic 4-week regimen of metformin reduces infarct
ischaemia-reperfusion injury. In murine models of size in rodents.6,7,9 Thus, it is unlikely that a longer
myocardial infarction, giving metformin, either before duration of metformin pretreatment would have
ischaemia or at reperfusion, consistently minimises resulted in clinically relevant protection. Third, a
infarct size.5 Data from human beings that show a general explanation for the absence of translation is
cardioprotective effect of metformin are scarce. In a that animal studies are predominantly done in young
retrospective analysis,18 metformin-treated patients with and otherwise healthy animals. By contrast, human
diabetes who had a myocardial infarction seemed to have patients with cardiovascular disease are older, and have
lower plasma troponin T concentrations than patients comorbidities and comedication, which can all interfere
given other antidiabetic drugs, but this study design is with cardioprotection.23 Among other factors, older age
prone to selection bias and confounding. In a randomised and insulin resistance, which were present in our
prospective pilot study19 in patients with metabolic patients, are associated with an impaired cardio-
syndrome, 1 week of treatment with metformin reduced protective effect of preconditioning, and drugs such
plasma high-sensitivity troponin I concentrations after as statins (used in most of the patients in our study)
elective percutaneous coronary intervention. This study, offer cardioprotective effects themselves.23–25 Propofol

www.thelancet.com/diabetes-endocrinology Published online July 13, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00121-7 7


anaesthesia can disrupt protection by remote ischaemic the first 24 h after surgery is a well established surrogate
preconditioning,4 but did not modulate the effect of marker of ischaemia-reperfusion injury, which is
metformin in our study. Additionally, publication bias independently associated with prognosis. Second, the
might exist in preclinical studies resulting in selective clinical secondary endpoints, such as length of stay in
publication of positive studies. the intensive-care unit, might be affected by factors
Animal studies have identified several signalling other than the clinical condition of the patient (eg,
cascades that contribute to metformin-induced logistical reasons), but this is unlikely to have affected
cardioprotection, including activation of myocardial our results because of the block-randomised, double-
AMP-activated protein kinase and Akt.7,8 Also, increased blind design. Third, we only measured plasma troponin I
intracellular formation of adenosine and subsequent and creatinine for 24 h after reperfusion. For troponin I,
adenosine receptor stimulation contribute to the direct a longer follow-up would not have improved the
cardioprotective effect of metformin.9 Finally, statistical power because troponin I was already low at
mitochondrial hexokinase II translocation increases 24 h. For creatinine, however, we might have missed the
resistance against ischaemia-reperfusion injury, and occurrence of acute kidney injury, because most CABG-
metformin can induce mitochondrial hexokinase II related acute kidney injury is detected only on day 2 after
translocation.26 In ex-vivo analyses of tissue from a surgery.30 Fourth, the exploratory laboratory analyses
subset of patients, we noted activation of AMP-activated were done, for practical reasons, on right atrial tissue,
protein kinase and Akt in atrial tissue, but to a smaller but effects might differ from the left ventricular tissue.
extent than previously reported in murine models.7,8 By Together with the results of randomised controlled
contrast, we did not find hexokinase II translocation. trials focusing on atherosclerosis and post-infarction
Notably, cardioprotective signalling pathways show remodelling,21,22 results from our study show that the
interspecies differences—e.g, Akt activation has been beneficial effects of metformin on atherosclerosis,
reported to be less important in ischaemic post- myocardial ischaemia-reperfusion injury, and post-
conditioning in larger mammals than in rodents.27 We infarction remodelling that were noted in preclinical
conclude that the mild activation of AMP-activated studies do not translate into a clinical benefit, at least in
protein kinase and Akt by metformin might not have patients without diabetes. Thus, the improved
been sufficient to offer significant cardioprotection. cardiovascular prognosis that has previously been
AMPD is a key intracellular enzyme catalysing the reported in patients with diabetes given metformin is
conversion of AMP into inosine monophosphate (IMP). likely to be due to effects on glucose metabolism and
The common loss-of-function variant 34C→T is bodyweight, rather than due to direct cardiovascular
associated with an impaired conversion of AMP into protective effects.
IMP, which facilitates the formation of adenosine, and is Contributors
associated with increased resistance against ischaemia- SEM, RN, CJZ, HAvS, PP, LN, H-JD, MJC, ARTD, AV, GAR, and NPR
reperfusion injury and better outcome in patients with acquired, analysed, and interpreted the data. SEM, RN, ARTD, GAR, and
NPR did the statistical analysis. SEM, RN, CJZ, GAR, and NPR drafted
cardiovascular disease.17,28 We postulated that in patients the manuscript. CJZ, HAvS, PP, LN, H-JD, GAR, and NPR supervised
with the variant allele, the metformin-induced increase the study, and HAvS, PP, GAR, and NPR conceived its design. MJC and
in adenosine formation would be augmented, and we did AV gave technical support, and GAR and NPR obtained funding. All
a prespecified subgroup analysis for the primary authors critically read and revised the manuscript.
endpoint in relation to this variant. This analysis, Declaration of interests
however, did not support our hypothesis; although our GAR has served on a scientific advisory board for Novartis. NPR has
received an unrestricted grant from AstraZeneca unrelated to this study,
study was powered to detect an effect of metformin in the and has served on a scientific advisory board for AstraZeneca. All other
patients with the CC genotype, it was not powered to authors declare no competing interests.
detect a larger effect of metformin in patients with the Acknowledgments
CT/TT genotype versus patients with the CC genotype. This study was supported by grant 2010B195 from the Netherlands Heart
Current practice is to temporarily stop giving Foundation to NPR and GAR. NPR is a recipient of a Dr Dekker grant
metformin before CABG surgery because of the from the Netherlands Heart Foundation (grant number 2012T51) and a
clinical fellowship from the Dutch Organisation for Health Research and
presumed increased risk of lactic acidosis. From the Development. RN is financially supported by a grant from the
scientific literature, however, evidence does not support Netherlands Heart Foundation (number NHS2010B011 awarded to CJZ).
this assumption, especially for patients with normal We thank the department of Cardiothoracic Surgery, Radboud University
Medical Center, Nijmegen, Netherlands, for their dedicated assistance.
kidney function.29 In our study, giving metformin to
We are particularly grateful to A Rasing-Hoogveld, V Verweij, and
patients with preserved renal function did not result in J Pertijs (Department of Pharmacology-Toxicology), and M Bouw
higher lactate concentrations, a higher occurrence of (Department of Intensive Care Medicine) for their contribution to the
lactate acidosis, or the occurrence of any serious collection of our data. We also thank P Van den Broek and A Bilos
(Department of Pharmacology-Toxicology) for their technical assistance
adverse event.
in measuring plasma caffeine and metformin concentrations, and
Our study has some limitations. First, it was not H van Outheusden (Jeroen Bosch Hospital, ‘s Hertogenbosch,
powered to detect differences in clinical endpoints. Netherlands) for his technical assistance in measuring plasma
However, the plasma troponin I concentration within high-sensitivity troponin I concentrations.

8 www.thelancet.com/diabetes-endocrinology Published online July 13, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00121-7


References 17 Hausenloy DJ, Mwamure PK, Venugopal V, et al. Effect of remote

1 Sipahi I, Akay MH, Dagdelen S, Blitz A, Alhan C. Coronary artery ischaemic preconditioning on myocardial injury in patients
bypass grafting vs percutaneous coronary intervention and undergoing coronary artery bypass graft surgery: a randomised
long-term mortality and morbidity in multivessel disease: controlled trial. Lancet 2007; 370: 575–79.
meta-analysis of randomized clinical trials of the arterial grafting 18 Lexis CP, Wieringa WG, Hiemstra B, et al. Chronic metformin
and stenting era. JAMA Intern Med 2014; 174: 223–30. treatment is associated with reduced myocardial infarct size in
2 Lurati Buse GA, Bolliger D, Seeberger E, et al. Troponin T and diabetic patients with ST-segment elevation myocardial infarction.
B-type natriuretic peptide after on-pump cardiac surgery: prognostic Cardiovasc Drugs Ther 2014; 28: 163–71.
impact on 12-month mortality and major cardiac events after 19 Li J, Xu JP, Zhao XZ, Sun XJ, Xu ZW, Song SJ. Protective effect of
adjustment for postoperative complications. Circulation 2014; metformin on myocardial injury in metabolic syndrome patients
130: 948–57. following percutaneous coronary intervention. Cardiology 2014;
3 Domanski MJ, Mahaffey K, Hasselblad V, et al. Association of 127: 133–39.
myocardial enzyme elevation and survival following coronary artery 20 El Messaoudi S, Schreuder TH, Kengen RD, et al. Impact of
bypass graft surgery. JAMA 2011; 305: 585–91. metformin on endothelial ischemia-reperfusion injury in humans
4 Heusch G. Cardioprotection: chances and challenges of its in vivo: a prospective randomized open, blinded-endpoint study.
translation to the clinic. Lancet 2013; 381: 166–75. PLoS One 2014; 9: e96062.
5 El Messaoudi S, Rongen GA, de Boer RA, Riksen NP. The 21 Preiss D, Lloyd SM, Ford I, et al. Metformin for non-diabetic patients
cardioprotective effects of metformin. Curr Opin Lipidol 2011; with coronary heart disease (the CAMERA study): a randomised
22: 445–53. controlled trial. Lancet Diabetes Endocrinol 2014; 2: 116–24.
6 Whittington HJ, Hall AR, McLaughlin CP, Hausenloy DJ, 22 Lexis CP, van der Horst IC, Lipsic E, et al. Effect of metformin on
Yellon DM, Mocanu MM. Chronic metformin associated left ventricular function after acute myocardial infarction in patients
cardioprotection against infarction: not just a glucose lowering without diabetes: the GIPS-III randomized clinical trial. JAMA
phenomenon. Cardiovasc Drugs Ther 2013; 27: 5–16. 2014; 311: 1526–35.
7 Calvert JW, Gundewar S, Jha S, et al. Acute metformin therapy 23 Ferdinandy P, Hausenloy DJ, Heusch G, Baxter GF, Schulz R.
confers cardioprotection against myocardial infarction via Interaction of risk factors, comorbidities, and comedications with
AMPK-eNOS-mediated signaling. Diabetes 2008; 57: 696–705. ischemia/reperfusion injury and cardioprotection by
8 Bhamra GS, Hausenloy DJ, Davidson SM, et al. Metformin protects preconditioning, postconditioning, and remote conditioning.
the ischemic heart by the Akt-mediated inhibition of mitochondrial Pharmacol Rev 2014; 66: 1142–74.
permeability transition pore opening. Basic Res Cardiol 2008; 24 Whittington HJ, Babu GG, Mocanu MM, Yellon DM,
103: 274–84. Hausenloy DJ. The diabetic heart: too sweet for its own good?
9 Paiva M, Riksen NP, Davidson SM, et al. Metformin prevents Cardiol Res Pract 2012; 2012: 845698.
myocardial reperfusion injury by activating the adenosine receptor. 25 van den Munckhof I, Riksen N, Seeger JP, et al. Aging attenuates
J Cardiovasc Pharmacol 2009; 53: 373–78. the protective effect of ischemic preconditioning against endothelial
10 Effect of intensive blood-glucose control with metformin on ischemia-reperfusion injury in humans.
complications in overweight patients with type 2 diabetes (UKPDS Am J Physiolog Heart Circ Physiol 2013; 304: H1727–32.
34). UK Prospective Diabetes Study (UKPDS) Group. Lancet 1998; 26 Nederlof R, Eerbeek O, Hollmann MW, Southworth R, Zuurbier CJ.
352: 854–65. Targeting hexokinase II to mitochondria to modulate energy
11 Hong J, Zhang Y, Lai S, et al. Effects of metformin versus glipizide metabolism and reduce ischaemia-reperfusion injury in heart.
on cardiovascular outcomes in patients with type 2 diabetes and Br J Pharmacol 2014; 171: 2067–79.
coronary artery disease. Diabetes Care 2013; 36: 1304–11. 27 Skyschally A, van Caster P, Boengler K, et al. Ischemic
12 Aldam P, Levy N, Hall GM. Perioperative management of diabetic postconditioning in pigs: no causal role for RISK activation. Circ Res
patients: new controversies. Br J Anaesth 2014; 113: 906–09. 2009; 104: 15–18.
13 Duncan AI, Koch CG, Xu M, et al. Recent metformin ingestion does 28 Anderson JL, Habashi J, Carlquist JF, et al. A common variant of
not increase in-hospital morbidity or mortality after cardiac surgery. the AMPD1 gene predicts improved cardiovascular survival in
Anesth Analg 2007; 104: 42–50. patients with coronary artery disease. J Am Coll Cardiol 2000;
14 Riksen NP, Franke B, Oyen WJ, et al. Augmented hyperaemia and 36: 1248–52.
reduced tissue injury in response to ischaemia in subjects with 29 Scale T, Harvey JN. Diabetes, metformin and lactic acidosis.
the 34C > T variant of the AMPD1 gene. Eur Heart J 2007; Clin Endocrinol 2011; 74: 191–96.
28: 1085–91. 30 Venugopal V, Laing CM, Ludman A, Yellon DM, Hausenloy D.
15 Wallace TM, Levy JC, Matthews DR. Use and abuse of HOMA Effect of remote ischemic preconditioning on acute kidney injury in
modeling. Diabetes Care 2004; 27: 1487–95. nondiabetic patients undergoing coronary artery bypass graft
16 Speechly-Dick ME1, Grover GJ, Yellon DM. Does ischemic surgery: a secondary analysis of 2 small randomized trials.
preconditioning in the human involve protein kinase C and the Am J Kidney Dis 2010; 56: 1043–49.
ATP-dependent K+ channel? Studies of contractile function after
simulated ischemia in an atrial in vitro model. Circ Res 1995;
77: 1030–35.

www.thelancet.com/diabetes-endocrinology Published online July 13, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00121-7 9