Hyperinsulinemic Normoglycemia during Cardiac Surgery reduces a Composite of 30-day Mortality and Serious In-hospital Complications: A Randomized Clinical Trial

Abstract

Background:

Hyperinsulinemic normoglycemia augments myocardial glucose uptake and utilization. We tested the hypothesis that hyperinsulinemic normoglycemia reduces 30-day mortality and morbidity after cardiac surgery.

Methods:

This dual-center, parallel group, superiority trial randomized cardiac surgical patients between August 2007 and March 2015 at the Cleveland Clinic and Royal Victoria Hospital to intraoperative glycemic management with: 1)hyperinsulinemic normoglycemia, a fixed high-dose insulin and concomitant variable glucose infusion titrated to glucose concentrations of 80–110 mg·dL⁻¹; or 2)standard glycemic management, low-dose insulin infusion targeting glucose >150 mg·dL⁻¹. The primary outcome was a composite of 30-day mortality, mechanical circulatory support, infection, renal or neurologic morbidity. Interim analyses were planned at each 12.5% enrollment of maximum 2,790 patients.

Results:

At the third interim analysis (N=1,439; hyperinsulinemic normoglycemia 709, standard glycemic management 730; 52% planned maximum), the efficacy boundary was crossed and study stopped per protocol. Time-weighted average glucose concentration (means±SDs) with hyperinsulinemic normoglycemia was 108±20 versus 150±33 mg·dL⁻¹ with standard glycemic management, P<0.001. At least one component of the composite outcome occurred in 49(6.9%) patients receiving hyperinsulinemic normoglycemia versus 82(11.2%) receiving standard glucose management (P<efficacy boundary 0.0085); estimated relative risk (95%interim-adjusted CI) 0.62(0.39,0.97), P=0.0043. There was a treatment-by-site interaction (P=0.063); relative risk for the composite outcome was 0.49(0.26,0.91, P=0.0007, N=921) at Royal Victoria Hospital, but 0.96(0.41,2.24, P=0.89, N=518) at the Cleveland Clinic. Severe hypoglycemia (<40 mg·dL⁻¹) occurred in 6(0.9%) patients.

Conclusions:

Intraoperative hyperinsulinemic normoglycemia reduced mortality and morbidity after cardiac surgery. Providing exogenous glucose while targeting normoglycemia may be preferable to simply normalizing glucose concentrations.

9. Summary Statement:

This clinical trial randomized 1,439 cardiac surgical patients to hyperinsulinemic normoglycemia versus standard glucose management and found that morbidity and mortality were reduced in patients who received hyperinsulinemic normoglycemia.

Introduction

Hyperglycemia is associated with mortality and morbidity in critically ill and cardiac surgical patients.^1–3 Consistent with these observations, intensive treatment of hyperglycemia aimed at normoglycemia reduced morbidity and mortality in a single center randomized trial of critically ill surgical patients, most of whom had recent cardiac surgery.⁴ Pediatric critically ill patients, most of whom had cardiac surgery and received intensive insulin therapy aimed at normoglycemia, similarly experienced reduced morbidity and mortality.⁵ Other trials, however, found that treatment of hyperglycemia with conventional insulin infusions aimed at normoglycemia either provided no benefit^6,7 or increased mortality.^8,9 Complications resulted, at least in part, from hypoglycemia.¹⁰

Disparities in reported outcomes may be related to whether or not sufficient glucose was provided. Outcomes in normoglycemic patients were generally favorable in trials where glucose was supplemented, either intravenously or nutritionally.^4,5,11,12 In contrast, outcomes were unfavorable when normoglycemia was produced only by insulin administration.^6–9 Provision of insulin and exogenous glucose while avoiding hyperglycemia, promotes myocardial glucose uptake and utilization, augments myocardial efficiency, and increases cardiac output^13–18 — all of which may improve outcomes by increasing systemic perfusion and end-organ function. Cardiac surgical patients may especially benefit from normoglycemia with supplemental glucose because intraoperative myocardial ischemia and reperfusion injury are common.^12,19,20 An additional benefit of normoglycemia is a reduced risk of perioperative infection.^21–23

Hyperinsulinemic normoglycemia is a well-established glycemic management technique in which exogenous glucose is combined with intensive insulin therapy to target normoglycemia.^24–26 Application of this technique in cardiac surgical patients aims to improve myocardial and end-organ function. Concurrent potassium supplementation is provided to avoid hypokalemia from insulin-induced cellular uptake of potassium.²⁷ The hyperglycemic normoglycemia technique thus bears a resemblance to glucose-insulin-potassium (GIK) therapy, ^12,28,29 except that normoglycemia is targeted. Normalization of glucose concentrations with the hyperinsulinemic normoglycemia technique may also reduce postoperative infections.

This investigation tested the hypothesis that intraoperative hyperinsulinemic normoglycemia improves a composite of 30-day postoperative mortality and serious cardiac, renal, neurologic, and infectious complications in patients recovering from cardiac surgery.

Materials and Methods

This dual-center, randomized, parallel-group, unblinded, superiority trial was approved by the Institutional Review Boards at the Cleveland Clinic, Cleveland, Ohio and Royal Victoria Hospital, Montreal, Quebec, and registered at ClinicalTrials.gov (NCT00524472) on August 31, 2007. Written, informed consent was obtained from each participant.

Adults between 18 and 90 years old scheduled for elective coronary artery bypass grafting (CABG), valve repair or replacement, or a combination of these procedures with cardiopulmonary bypass between August 2007 and April 2015 were screened for inclusion by research personnel. Exclusion criteria included off-pump cardiac surgery, anticipated hypothermic circulatory arrest, elevated baseline cardiac troponin I (>0.5 ng·l⁻¹, Montreal) or troponin T (>0.1 ng·ml⁻¹, Cleveland), kidney disease requiring renal replacement therapy, or active infection requiring ongoing antibiotic therapy. Sub-investigations, unrelated to the primary outcome, have previously been published.^19,30–35

Randomization and masking

Study participants were randomly assigned (1:1) to hyperinsulinemic normoglycemia or standard glycemic management. Randomization was performed by the Plan procedure in SAS software, a web-based system, and was stratified by center (Cleveland versus Montreal), cardiac surgical procedure (coronary artery bypass grafting, valve repair/replacement, or combined procedure) and history of diabetes (type 2/diet-controlled versus type 1). Block size within each stratum randomly ranged from four to 16 patients. Allocation was initially concealed in sealed, sequentially numbered envelopes, and later in a web-based system, both accessed shortly before induction of anesthesia.

It was not feasible to blind anesthesia and surgical personnel to the intraoperative glucose management strategy; however, primary outcomes and postoperative clinical and laboratory results were evaluated by research personnel blinded to group allocation.

Procedures

Anesthesia and Surgery

Standard anesthesia monitors were supplemented by central venous or pulmonary artery catheters and transesophageal echocardiography. Midazolam, etomidate, thiopental, propofol, sufentanil and/or fentanyl, volatile anesthetics, and a depolarizing or non-depolarizing muscle relaxant were given during induction and maintenance of anesthesia. Surgery was performed through a full midline sternotomy or minimally invasive upper hemisternotomy, and routine strategies for conduct of cardiopulmonary bypass were followed.

Intermittent antegrade and retrograde administration of Buckberg’s cardioplegia mixed in 5% dextrose was used exclusively in Cleveland until December 2012; thereafter del Nido cardioplegia, a non-glucose containing solution administered as a single anterograde infusion, was occasionally used for isolated valve repair/replacement without CABG. In Montreal, intermittent anterograde and/or retrograde St. Thomas cardioplegia, a non-glucose containing solution was administered. Intravenous vasoactive infusions and antibiotic medications were mixed in 5% dextrose in Cleveland and normal saline solution in Montreal.

During separation from cardiopulmonary bypass, epinephrine was infused for low cardiac index (<2.0 L∙min⁻¹∙m⁻²) and/or norepinephrine or vasopressin were infused for low systemic vascular resistance (<700 dyn∙sec∙cm⁻⁵) to maintain mean arterial pressure >80 mmHg and cardiac index >2.0 L∙min⁻¹∙m⁻². Milrinone was infused when cardiac output was low and refractory to routine pharmacologic hemodynamic support. If a pulmonary artery catheter was not present, transesophageal echocardiography was used to assess myocardial contractility and determine whether inotropic versus vasopressor support was needed.

Glucose management

Intraoperative glucose management with hyperinsulinemic normoglycemia involved a fixed-dose insulin infusion of 5 mU·kg^−1.min⁻¹ with a concomitant variable glucose (dextrose 20%) infusion supplemented with potassium (40 mEq·L⁻¹) and phosphate (30 mmol·L⁻¹) as previously described.²⁴ The glucose infusion was initiated at approximately 40–60 mL·hr⁻¹ when serum glucose concentration was approximately 110 mg·dL⁻¹ or less, and manually titrated to target glucose concentrations of 80–110 mg·dL⁻¹ every 10 to 15 minutes throughout surgery. Additional boluses of insulin were given for blood glucose >110 mg·dL⁻¹. Arterial blood glucose concentrations were measured with an Accu-Check (Roche Diagnostics, Switzerland) glucose monitor. At sternal closure, the insulin infusion was reduced to 1 mU·kg^−1.min⁻¹ and converted to a standard low-dose insulin infusion upon Intensive Care Unit (ICU) admission. After ICU arrival, the glucose infusion was decreased by 25 – 50% every 20 min when the blood glucose was >110 mg·dL⁻¹. When the infusion was at 20 cc·hr⁻¹ or less and blood glucose was >110 mg·dL⁻¹, the infusion was discontinued. Blood glucose concentrations were followed for 45 – 60 min after discontinuation of the dextrose infusion to ensure that hypoglycemia was avoided.

Standard glucose management involved a conventional low-dose insulin infusion titrated to blood glucose concentrations measured by arterial blood gas analysis every 30 – 90 minutes throughout surgery. This low-dose insulin infusion was initiated for blood glucose concentration >120 mg·dL⁻¹ before initiation of cardiopulmonary bypass or >150 mg·dL⁻¹ during or after cardiopulmonary bypass, at a rate based on patient weight and current glucose concentration. Subsequent adjustments were based on a sliding scale of current blood glucose concentration and the change from the previous measurement. Supplemental boluses of insulin were given with acute increases (>30 mg·dL⁻¹) in blood glucose. The insulin protocol for patients assigned to standard glucose management is listed in Appendix 1.

Upon ICU admission, both groups transitioned to the same standardized postoperative insulin treatment protocol in the ICU. This involved measurement of blood glucose by arterial blood gas analysis approximately every two hours with adjustment of insulin infusion to maintain serum glucose <150 mg·dL⁻¹ on postoperative day one and <120 mg·dL⁻¹ on day two and later. In 2009 following publication of the NICE-SUGAR trial⁹, the postoperative glucose target increased to <180 mg/dL.

Severe and moderate hypoglycemia was defined as blood glucose less than 40 and 60 mg·dL⁻¹, respectively. Hypoglycemia was treated by administration of 20% dextrose (25100 ml). A summary of major protocol changes that occurred since initiation of this investigation are found in Appendix 2.

Outcomes

The primary outcome was a collapsed composite (any versus none) of the following major postoperative complications occurring within 30 days of surgery, including 1) all-cause postoperative mortality; 2) failure to wean from cardiopulmonary bypass or postoperative low cardiac index (<1.8 L·min^−1.m⁻²) requiring mechanical circulatory support with intraaortic balloon counter-pulsation, ventricular assist device, and/or extracorporeal mechanical oxygenation; 3) serious postoperative infection including any of the following infectious complications: mediastinitis, sternal wound infection requiring surgical debridement, sepsis, or pneumonia requiring mechanical ventilatory support; 4) acute postoperative kidney injury requiring renal replacement therapy; and 5) new postoperative focal (aphasia, decrease in limb function, hemiparesis) or global (diffuse encephalopathy with >24 hours of severely altered mental status or failure to awaken postoperatively) neurologic deficit.

The secondary outcomes included postoperative atrial fibrillation, defined as the occurrence of new onset postoperative atrial fibrillation following cardiac surgery, duration of hospitalization (days) and ICU stay (days), and one year all-cause mortality.

We also recorded a composite of minor postoperative complications within the first 30 days including mechanical ventilation >72 hours, low cardiac index (cardiac index <1.8 l·min^−1.m⁻² despite adequate fluid replacement (lack of hemodynamic response to repeated fluid administration of crystalloid or colloid intravascular solutions) and high dose inotropic support for >4 hours), acute kidney injury (increase in creatinine >100%), hospitalization >30 days, and all-cause hospital readmission within 30 days. Detailed definitions of the primary and secondary outcomes are listed in Appendix 3.

Statistical analysis

Balance on baseline characteristics between randomized groups was assessed using the standardized difference (STD = difference in means or proportions divided by pooled standard deviation). Imbalance was defined as a STD >0.2 in absolute value,³⁶ and such variables were adjusted for in all analyses.

We assessed the effect of hyperinsulinemic normoglycemia versus standard therapy on the primary outcome (any complication) using Cochran-Mantel-Haenszel chi-square analysis, adjusting for clinical site. Results were reported as the estimated relative risk and interim analysis adjusted 95% confidence interval. We assessed the interaction between treatment effect and site using the Breslow-Day test for homogeneity of odds ratios. We also assessed the treatment-by-component interaction overall and within site using multivariate (one record per component per subject) generalized estimating equation “distinct-effects” logistic models.³⁷

Groups were compared on binary secondary outcomes using the same methods as for the primary outcome. The treatment effect on time-to-event secondary outcomes (i.e., duration of mechanical ventilation, ICU and hospital stay [time to discharge alive]) were assessed using Cox proportional hazards models adjusting for site. For patients who died during the index hospitalization (N = 22), the hospital stay was assigned to be the longest observed hospital stay + 1 day, and censored at that time (i.e., not discharged alive). “Time to discharge alive” was not done used for ICU length of stay because the exact date/time of death was not recorded (just whether in-hospital or not). Median (95% CI) survival time was estimated from Kaplan-Meier curves.

Intraoperative time-weighted mean glucose concentration was calculated across measurements for each patient using the trapezoidal method and equal to the area under the curve divided by the total glucose reading time.

The significance level for each hypothesis was 0.05 and all tests were 2-sided. Confidence intervals were adjusted for the group sequential design (using confidence coefficient of 2.63) to maintain overall study alpha of 0.05 for combined sites and 0.025 overall (confidence coefficient of 2.86) within sites. Significance criterion for treatment-by-site interaction was set at 0.10 a priori. Bonferroni correction was performed while assessing each individual component of the composite primary outcome, with the significant level of 0.0017 (i.e., 0.0085/5 components = 0.0017) with 99.83% confidence intervals (CI), and 0.00084 (i.e., 0.0042/5) with 99.92% CI within site. SAS 9.2 (Carey, NC) or East 5.3 (Cytel Corporation) were used for all analyses.

Sample size calculations

A maximum of 2,790 patients was required to detect a 30% relative reduction in the composite of any major complications (i.e. any versus none) from an expected 15% incidence of complications in the standard group at the overall 0.05 significance level with 90% power. Interim analyses to assess efficacy and futility on the primary outcome of the occurrence of any major complication were planned at each 12.5% of the maximum planned enrolment in this group sequential design (N = 349, 697, 1046, 1394, 1743, 2091, and 2440). Patient recruitment continued while the interim analyses were performed, thus the timing of the interim analyses varied slightly from the original plan. We used the alpha (Type I error) and beta (Type II error) spending approach of Hwang et al.³⁸ with parameters gamma = −2 for efficacy and gamma = −3 for futility.

Results

Patients were recruited from August 17, 2007 until March 30, 2015; 1,439 patients were randomly assigned to hyperinsulinemic normoglycemia (N = 709) and standard glycemic management (N = 730), with 518 in Cleveland and 931 in Montreal (Figure 1). The number of patients that were screened for this investigation was not available. At the third interim analysis with N = 1,439 (52% of maximum enrolment; patient recruitment continued during data analysis, thus the third interim analysis was later than initially planned), the treatment effect of hyperinsulinemic normoglycemia on the primary outcome crossed the predefined efficacy boundary for the combined sites, and the study was stopped as per the protocol. The P value boundaries for efficacy and futility were P <0.0085 and P ≥ 0.803, respectively.

Figure 1.

Patient flow diagram

Randomized groups were well-balanced (absolute standardized difference <0.20) on all preoperative patient demographics, clinical characteristics, preoperative echocardiographic measurements, and perioperative variables (Table 1). One year survival data were unavailable on 104 patients (hyperinsulinemic normoglycemia N = 56; standard glycemic management N = 48).

Table 1.

Baseline and surgical characteristics by treatment and site. Data are presented as N (%), Mean ± SD or Median [1^st, 3^rd quartiles]

	Combined Sites			Cleveland Clinic			Royal Victoria Hospital
Factor	Hyperinsulinemic Normoglycemia (N=709)	Standard Therapy (N=730)	ASD	Hyperinsulinemic Normoglycemia (N=252)	Standard Therapy (N=266)	ASD	Hyperinsulinemic Normoglycemia (N=457)	Standard Therapy (N=464)	ASD
Demographics
Age	66 ± 11c	66 ± 11b	0.02	66 ± 13a	66 ± 12a	−0.03	67 ± 11c	66 ± 10b	0.05
Female	189 (27)b	184 (26)a	0.04	76 (30)a	72 (27)a	0.07	113 (26)b	112 (25)a	0.02
White race	596 (87)b	615 (86)a	0.01	245 (97)a	248 (93)a	0.19	351 (81)b	367 (82)a	−0.04
Body mass index	28.5 ± 5.7b	28.3 ± 5.4a	0.03	29.6 ± 6.1a	28.9 ± 5.7a	0.11	27.9 ± 5.4b	28.0 ± 5.2a	−0.01
Medical History
Diabetes	226 (32)a	249 (34)a	−0.05	69 (27)a	74 (28)a	−0.01	157 (34)a	175 (38)a	−0.07
COPD/Asthma	107 (16)b	85 (12)a	0.10	41 (16)a	34 (13)a	0.10	66 (15)b	51 (11)a	0.11
Pulmonary Hypertension	101 (15)b	102 (14)a	0.01	57 (23)a	66 (25)a	−0.05	44 (10)b	36 (8)a	0.07
Stroke	41 (6)b	32 (5)b	0.07	23 (9)a	19 (7)a	0.07	18 (4)b	13 (3)b	0.07
Hypertension	533 (77)b	561 (79)a	0.03	160 (63)a	171 (64)a	0.02	373 (86)b	390 (87)a	0.04
Heart failure	146 (21)b	137 (19)a	0.05	77 (31)a	73 (27)a	0.07	69 (16)b	64 (14)a	0.047
Myocardial infarction	193 (28)b	173 (24)a	0.09	59 (23)a	62 (23)a	0.002	134 (31)b	111 (25)a	0.13
Dialysis	4 (1)b	4 (1)a	0.003	2 (1)a	3 (1)a	0.03	2 (0)b	1 (0)a	0.04
Peripheral vascular disease	51 (7)b	42 (6)a	0.06	32 (13)a	28 (11)a	0.07	19 (4)b	14 (3)a	0.06
Smoking	197 (29)b	180 (25)a	0.05	109 (43)a	117 (44)a	0.01	88 (20)b	63 (14)a	0.16
ASA physical status			0.08			0.14			0.02
2	2 (0)b	0 (0)a		2 (1)a	0 (0)a		0(0)a	0(0)a
3	334 (49)	349 (49)		46 (18)	48 (18)		288 (66)	301 (67)
4	348 (51)	363 (51)		201 (80)	215 (81)		147 (34)	148 (33)
5	2 (0)	1 (0)		2 (1)	1 (0)		0(0)	0(0)
Preoperative medications
ACE Inhibitor	266 (39)b	262 (37)b	0.03	96 (38)a	96 (36)a	0.04	170 (39)b	166 (38)b	0.03
Antiarrhythmic	56 (8)b	67 (10)b	0.05	34 (13)a	44 (17)a	0.09	22 (5)b	23 (5)b	0.01
Beta-blocker	434 (63)b	452 (64)a	0.01	116 (46)a	134 (50)a	0.09	318 (73)b	318 (71)a	0.03
Calcium Blocker	128 (19)b	147 (21)b	0.05	32 (13)a	47 (18)a	0.14	96 (22)b	100 (23)b	0.01
Cox-2 Inhibitor	13 (2)d	3 (0)d	0.14	6 (3)c	1 (0)c	0.19	7 (2)b	2 (0)b	0.12
Statin	475 (69)b	486 (69)b	0.004	130 (52)a	145 (55)a	0.06	345 (79)b	341 (78)b	0.04
Steroid	33 (5)b	23 (3)b	0.08	11 (4)a	16 (6)a	0.07	22 (5)b	7 (2)b	0.20
Anti-diabetic drugs	145 (21)b	144 (20)a	0.02	6 (2)a	6 (2)a	0.008	139 (32)b	138 (31)a	0.02
Sulfonylureas or Meglitinides	52 (8)d	58 (9)d	0.02	21 (10)c	18 (8)c	0.071	31 (7)b	40 (9)b	0.07
Biguanides (metformin)	126 (20)d	124 (19)d	0.02	24 (12)c	29 (13)c	0.05	102 (23)b	95 (21)b	0.05
Thiazolidinediones	16 (3)d	11 (2)d	0.06	7 (3)c	4 (2)c	0.10	9 (2)b	7 (2)b	0.04
Insulin	70 (11)d	68 (10)d	0.02	24 (12)c	26 (12)c	0.005	46 (11)b	42 (9)b	0.04
Preoperative echocardiographie measurements
Mitral regurgitation severity			0.17			0.25			0.29
0	97 (27)g	89 (24)c		74 (37)c	77 (36)c	0.04	23 (14)f	12 (8)g
+ 1	85 (23)	86 (23)		42 (21)	39 (18)	0.09	43 (27)	47 (31)
+2	78 (22)	66 (18)		34 (17)	23 (11)	0.09	44 (28)	43 (28)
+3	46 (13)	47 (13)		21 (10)	31 (14)		25 (16)	16 (11)
+ 4	56 (15)	79 (22)		31 (15)	46 (21)		25 (16)	33 (22)
Mitral stenosis	8 (2)g	6 (2)c	0.04	6 (3)c	5 (2)c	0.04	170 (39)b	1 (1)g	0.05
Aortic regurgitation severity			0.18			0.17			0.36
0	187 (48)g	194 (51)c		142 (57)a	164 (62)a		45 (33)h	30 (26)h
+1	77 (20)	91 (24)		42 (17)	42 (16)		35 (25)	49 (42)
+2	66 (17)	46 (12)		38 (15)	28 (11)		28 (20)	18 (16)
+3	32 (8)	24 (6)		17 (7)	14 (5)		15 (11)	10 (9)
+4	27 (7)	25 (7)		12 (5)	16 (6)		15 (11)	9 (8)
Aortic stenosis	225 (47)b	201 (43)b	0.08	119 (47)a	111 (42)a	0.11	106 (47)b	90 (45)b	0.04
LV ejection fraction			0.13			0.07			0.20
LVEF > 60%	246 (37)c	257 (37)c		89 (39)b	91 (36)a		157 (36)b	166 (38)b
LVEF 50 – 59%	236 (36)	238 (34)		91 (39)	106 (42)		145(34)	132 (30)
LVEF 45 – 49%	3 (0)	3 (0)		3 (1)	3 (1)
LVEF 40 – 44%	54 (8)	65 (9)		10 (4)	13 (5)		44 (10)	52 (12)
LVEF 35 – 39%	34 (5)	54 (8)		11 (5)	12 (5)		23 (5)	42 (10)
LVEF <35%	89 (13)	76 (11)		27 (12)	28 (11)		62(14)	48 (11)
Preoperative laboratory measurements
BUN (mg/dL)	20.9 ± 16.7e	21.0 ± 17.1e	0.006	22 ± 11a	21.5 ± 10.2a	0.05	20 ± 20e	21 ± 21e	0.03
Creatinine (mg/dL)	96.7 ± 45.6b	97.1 ± 52.5b	0.007	104 ± 66a	101 ± 62a	0.05	92 ± 27b	95 ± 46b	0.06
Hematocrit (mg/dL)	40 ± 8b	40 ± 8b	0.03	40 ± 5a	41 ± 5a	0.09	40 ± 9b	40 ± 10b	0.007
Perioperative variables
Cardiac surgical procedure CABG (No valve)	298 (42)	311 (43)	0.02	85 (34)	87 (33)	0.05	213 (47)	224 (48)	0.03
Valve (No CABG)	272 (38)	281 (38)		97 (38)	109 (41)		175 (38)	172 (37)
CABG + Valve	138 (19)	138 (19)		70 (28)	70 (26)		68 (15)	68 (15)
Previous Cardiac Surgery	94 (14)b	84 (12)a	0.06	69 (27)	68 (26)	0.04	25 (6)b	16 (4) a	0.10
Surgical variables
Duration of surgery (min)	260 [200, 341]f	273 [210, 351]f	0.08	376 [311, 444]c	358 [305, 423]c	0.1	215 [170, 250]e	220 [180, 270]e	0.14
Duration of cardiopulmonary bypass (min)	96 [77, 126]d	99 [77, 126]d	0.01	93 [77, 120]c	94 [74, 116]c	0.04	99 [77, 130]b	101 [79, 130]b	0.04
Duration of aortic cross-clamp (min)	77 [60, 104]d	79 [61, 103]d	0.02	71 [58, 93]c	74 [57, 93]c	0.001	82 [61, 108]b	84 [64, 107]b	0.03
Cardioplegia			0.07			0.12			0
St. Thomas	457 (64)	464 (64)		0 (0)	0(0)		457 (100)	464 (100)
Buckberg’s	246 (35)	257 (35)		246 (98)	257 (97)		0(0)	0 (0)
Del Nido	5 (1)	9 (1)		5 (2)	9 (3)		0(0)	0 (0)
Microplegia	1 (0)	0 (0)		1 (0)	0 (0)		0(0)	0(0)

Data are represented as N (%), mean ± SD, or median [25^th, 75^th percentiles]. CC = Cleveland Clinic; RVH = Royal Victoria Hospital; COPD = Chronic obstructive pulmonary disease; ASA = American Society of Anesthesiologists (ASA) physical status classification system; ACE inhibitor: Angiotensin converting enzyme inhibitor; LVEF = Left ventricular ejection fraction; CABG = Coronary artery bypass grafting. LV = Left ventricular tASD = Absolute standardized difference (difference in means or proportions divided by standard deviation); imbalance defined as absolute value of ASD > 0.20 (small effect size).

Missing data points:

^a1 ~19

^b20 ~29

^c35~51

^d70~79

^e150~162

^f 190~200

^g231~299

^h300~514.

Insulin administration and treatment effects

Overall mean ± SD time-weighted average glucose concentration was 108 ± 20 mg·dL⁻¹ with hyperinsulinemic normoglycemia versus 150 ± 33 mg·dL⁻¹ with standard glycemic management. The Cleveland site had higher time-weighted average glucose concentration than Montreal overall, as well as within each treatment (all P<0.001). In Cleveland, glucose concentration was 121 ± 19 mg·dL⁻¹ with hyperinsulinemic normoglycemia and 171 ± 31 mg·dL⁻¹ with standard glycemic management. At the Royal Victoria Hospital, patients in the hyperinsulinemic normoglycemia group had glucose concentrations of 101 ± 17 mg·dL⁻¹ versus 136 ± 26 mg·dL⁻¹ with standard glycemic management. Reduction in mean time-weighted average intraoperative glucose concentration was similar at each site, with the estimated ratio of means (hyperinsulinemic normoglycemia/standard) (95% CI) being 0.71 (0.69, 0.73) in Cleveland versus 0.74 (0.72, 0.76) in Montreal. The overall effect for combined sites was 0.73 (0.72, 0.74; Figure 2).

Figure 2.

Boxplots comparing randomized groups on time weighted intraoperative glucose concentrations overall and within site. All = combined sites; HN = hyperinsulinemic normoglycemia. Box shows the interquartile range; horizontal line marks the median; whiskers extend to high and low values within 1.5 interquartile range of the box; circles are values beyond 1.5 interquartile range of the box; diamond shows the mean.

Moderate hypoglycemia (glucose concentration <60 mg·dL⁻¹) occurred in 91 (13%) of the hyperinsulinemic normoglycemia group and severe hypoglycemia (<40 mg·dL⁻¹) occurred in 6 (0.9%). The average duration of a hypoglycemic episode in the hyperinsulinemic normoglycemic group was 9 (range 3–16) minutes. Only one patient in the conventional insulin infusion group had severe hypoglycemia, lasting 29 minutes.

Primary results

At least one component of the composite outcome occurred in 49 (6.9%) of patients receiving hyperinsulinemic normoglycemia versus 82 (11.2%) receiving standard glucose management (P< efficacy boundary of 0.0085) for an estimated relative risk (95% interim-adjusted CI) of 0.62 (0.39, 0.97), P = 0.0043 (Figure 3a). However, there was a strong treatment-by-site interaction (P = 0.063, less than the a priori criterion of 0.10); the relative risk for the composite outcome was 0.49 (0.26, 0.91, P = 0.0007, N = 921) at the Royal Victoria Hospital, Montreal, but 0.96(0.41, 2.24, P = 0.89, N = 518) at the Cleveland Clinic. Proportions and relative risks for the individual major complications in the combined sites are shown in Figure 3a and by individual site in Figure 3b. There was no evidence of treatment-by component interaction overall (P=0.84), for Cleveland (P = 0.96) or Montreal (P = 0.52), and thus inference within components was statistically unnecessary. Nevertheless, after adjusting for multiple comparisons across components, only serious infection morbidity in Montreal was significantly affected by intervention.

Figure 3.

Comparison of the hyperinsulinemic normoglycemia (HN) and standard therapy group on the composite outcome of any major morbidity/30-day mortality and individual components of the composite outcome at combined sites (Figure 3a) and within individual sites (Figure 3b). CI=confidence interval; Interaction P-value (treatment-by-site) = 0.063. Confidence intervals adjusted for group sequential design (using confidence coefficient of 2.633) to maintain overall study alpha of 0.05 for combined sites and confidence coefficient of 2.86 within sites. P-values for combined sites: significant if P < 0.0085 for efficacy (with 99.15% CI); P-values for each site: significant if P < 0.0042 for efficacy (with 99.58% CI); P-values for each component: significant if P < 0.0042/5 = 0.00084 (with 99.92% CI) using Bonferroni correction.

Secondary results

Time-to-event secondary outcomes for the combined sites are shown in Table 2; no differences were found on any of the five outcomes. Secondary outcomes are shown by site in Appendix 4; no differences were found in any secondary outcome within site. Although there was a significant treatment-by-site interaction for ICU length of stay (P = 0.046), the treatment effect was not significant for either site; the hyperinsulinemic normoglycemia group was an estimated 1.24 (0.98, 1.57) times more likely to be discharged earlier than in the standard group in Montreal (P = 0.0026, not significant after Bonferroni correction), and 0.99 (0.74, 1.33) at the Cleveland Clinic (P = 0.89). Similarly, the treatment-by-site interaction was significant for hospital stay (P = 0.07), but the treatment effect was not significant at either site. There were no differences between groups on other secondary outcomes, including the composite of minor complications, postoperative atrial fibrillation, or one-year mortality (i.e. all P >0.0085, Table 2).

Table 2.

Comparison of the Hyperinsulinemic Normoglycemia vs. Standard Therapy groups on Secondary Outcomes

Outcomes	Hyperinsulinemic Normoglycemia (N = 709)		Standard Therapy (N = 730)
	N	Median (95% CI)	N	Median (95%CI)	HR (99.83%CI)*	P-value*
ICU stay (hrs)						0.046b
	649	25(24.9, 26.3)	671	27 (25.2, 27.3)	1.13 (0.95, 1.35)	0.025
Hospital stay (days)**						0.07 b
	686	8 [6, 12]	713	8 [6, 12]	1.05 (0.89, 1.25)	0.35
	N	Event (%)	N	Event (%)	RR (99.83%CI)*
Postoperative atrial fibrillation						0.51 b
	709	209 (29)	730	235 (32)	0.92 (0.75,1.13)	0.29
Any minor complicationa						0.21 b
	709	194(27)	730	227(31)	0.95 (0.87,1.04)	0.13
1-year mortality						0.13 b
	653	32 (5)	682	22 (3)	1.52 (0.74,3.11)	0.12

^*Confidence intervals and P values from Cox proportional hazards for ICU stay and Hospital stay, and Cochran-Mantel-Haenszel test for binary outcomes. Confidence intervals adjusted for group sequential design using confidence coefficient of 2.633 for combined sites in order to maintain overall study alpha of 0.05. Significant if P < 0.0085 /6 = 0.0017 using Bonferroni correction.

^**The observed longest hospital stay + 1 day was assigned to patients who died during hospitalization (N = 22).

^aMinor complications include any one of the following: a prolonged requirement for mechanical ventilation (>72 hours), low cardiac index (cardiac index <1.8 l/min/m² despite adequate fluid replacement and high dose inotropic support for >4 hours), acute kidney injury (increase in creatinine >100%), prolonged hospitalization (>30 days), and all-cause hospital readmission within 30 days).

^btreatment × site P-value.

Discussion

Hyperinsulinemic normoglycemia reduced the composite outcome of 30-day mortality and serious complications by nearly 40% (confidence interval 3–61%) in patients having cardiac surgery across our two clinical sites. Our results are broadly consistent with previous work showing that normoglycemia reduces various complications when supplemental glucose is provided.^4,5,11,12

The fixed high-dose insulin infusion technique contrasts with most previous trials in which only insulin was given to maintain normoglycemia. Insulin is cardioprotective independent of glucose concentrations.³⁹ Insulin administration during reperfusion reduces myocardial infarction via Akt and p70s6 kinase-dependent signaling pathways^28,39 and may improve myocardial metabolic and functional recovery after cardioplegic arrest.^40,41 Laboratory investigations similarly report myocardial benefit from provision of glucose and insulin.⁴²

The protective effects of enhanced myocardial glucose uptake and utilization may be especially beneficial during cardiac surgery because it might counteract myocardial dysfunction consequent to cardioplegic arrest and ischemia and reperfusion injury. Previous studies that largely enrolled cardiac surgical patients (N = 1548 and 700) similarly demonstrated benefit, although intensive insulin therapy targeting normoglycemia with supplemental glucose was initiated following surgery.^4,5 One other investigation (N = 371), however, examined the benefit of intraoperative glucose control during cardiac surgery and reported worse outcomes with intensive insulin therapy,⁸ although a standard insulin infusion, rather than hyperinsulinemic normoglycemia, was evaluated.

Hyperinsulinemic normoglycemia resembles GIK therapy which provided myocardial protection and improved left ventricular function in some,^14,28 but not all,⁴³ investigations. Both approaches are thought to provide cardioprotective benefits by increasing myocardial glucose uptake and improving coupling of glycolysis and glucose utilization.^42,44,45 However, hyperinsulinemic normoglycemia differs from GIK in avoiding hyperglycemia which is consistently associated with worse outcomes.^3,46 Variable degrees of hyperglycemia may explain why GIK demonstrated benefit in some investigations^12,28 but not in others.^29,47,48

Aside from the overall significant benefit of hyperinsulinemic normoglycemia, the most striking aspect of our results is that the benefit was apparently restricted to one study site. We considered several potential explanations. Although glycemic management was standardized, cardioplegia at the Cleveland Clinic contained glucose whereas it did not in Montreal; thus both groups at the Cleveland Clinic received exogenous glucose during cardioplegic arrest. It is possible that the usual provision of glucose-containing cardioplegia provided significant myocardial protection and reduced low cardiac output syndrome and mechanical circulatory support in all patients at the Cleveland Clinic, regardless of randomized group.

The need for mechanical circulatory support was low (<2%) in all groups that received glucose from either hyperinsulinemic normoglycemia or cardioplegia administration. Only the standard glycemic management group in Montreal did not receive exogenous glucose during cardioplegic arrest and also demonstrated the highest need for mechanical circulatory support. Consistent with this theory, a previously reported sub-investigation⁴⁹ from Montreal provided evidence of cardio-protection and improved myocardial function in patients who received hyperinsulinemic normoglycemia, but not standard glucose management. In contrast, myocardial function at the Cleveland Clinic was not different between groups.³⁰

Glucose concentrations for both randomized groups were higher in Cleveland than in Montreal, presumably because patients at the Cleveland Clinic were given cardioplegia with glucose and medications were mixed with glucose. Higher glucose concentrations at the Cleveland Clinic may explain the lack of difference between groups, whereas the effect was profound at the Royal Victoria Hospital.⁵⁰ Results for the primary outcome were clearly centered around the null hypothesis at the Cleveland Clinic, with a relative risk estimate of 0.96. However, because Cleveland contributed only about a third of the patients, the site-specific 95% confidence intervals for the primary outcome range from a 59% reduction to a 2.2-fold increase in the composite outcome, which does not allow a firm negative conclusion.

Hyperinsulinemic normoglycemia reduced serious postoperative infection only in Montreal. Others similarly reported a nearly 50% reduction in blood stream and sternal wound infections in cardiac surgical and critically ill patients who received intensive insulin therapy.^4,22 Hyperglycemia impairs leukocyte function, increasing risk of infection,^51,52 and our results are consistent with these observation. The Cleveland site however, received no benefit from hyperinsulinemic normoglycemia. The incidence of postoperative infectious complications in the Cleveland control group was half of the incidence of infection in Montreal. It is therefore possible that infection risk at the Cleveland Clinic was already low so that hyperinsulinemic normoglycemia provided little additional benefit.

Hypoglycemia, which has been closely linked to adverse outcomes in other investigations, rarely occurred in our study. We attribute the low incidence of hypoglycemia to the profound stress counter-regulatory response and insulin resistant state that ensues with cardiac surgery and during the conduct of cardiopulmonary bypass. But it is also due to frequent blood glucose measurements (generally every 10–15 min) and close titration of glucose by dedicated investigators.

We could not blind anesthesia or surgical personnel to intraoperative glycemic management; however, most outcomes occurred several hours to days postoperatively and were recorded by research personnel who were blinded to treatment assignment. Our investigation cannot determine whether the benefit of hyperinsulinemic normoglycemia was due to the administration of high-dose insulin with glucose supplementation versus benefits of normoglycemia; thus the observed benefits may have resulted from more intensive glucose, rather than the concomitant provision of supplemental glucose. The study stopped after slightly more than 50% of the planned patients were enrolled, but it was not “stopped early” for logistical or other non-statistical reasons; enrollment was stopped per protocol by the Executive Committee because results at a planned interim analysis met a priori efficacy criteria. Our “group sequential” design protected the type I error at 5% and the type II error at 10% for the primary analyses. That said, as is true with any such design which crosses a boundary and thus (legitimately) stops enrollment before the maximum is reached, our confidence intervals would have been somewhat narrower had we continued.

In summary, hyperinsulinemic normoglycemia in patients having cardiac surgery reduced a composite of postoperative morbidity and mortality. Because previous investigations targeting normoglycemia in the absence of exogenous glucose supply found no benefit, targeting normoglycemia while providing exogenous glucose may be preferable to simply normalizing blood glucose concentrations.

Appendix 1. Cleveland Clinic Operating Room Insulin Therapy Protocol

Blood Glucose Goal: 70 – 150 mg/dL. Regular Insulin 100 units/100 ml in 0.9% normal saline in a concentration of 1 unit/ml will be used.

1. Starting Insulin: Start if pre CPB blood glucose > 120, and if on pump or post pump blood glucose > 150.

   • Bolus dose: 0.03 units/kg (maximum bolus is 3 units)
   • Initiate continuous infusion: initial rate 0.03 units/kg/hr (maximum initial rate is 3 units/hr)
   • See Table 1 (Insulin Infusion adjustment) for adjustment of insulin rate.

2. Blood glucose monitoring: Measure blood glucose between 30 – 60 minutes during surgery. (This recommendation was changed to 60 – 90 min in 2009).

3. Hypoglycemia protocol:

   • If blood glucose ≤ 60 mg/dL: stop insulin infusion, give 25 – 50 mL of 50% dextrose solution, obtain blood glucose level every 30 minutes until blood glucose > 80 mg/dL for three consecutive levels, and then check blood glucose every 30 – 60 minutes.
   • If blood glucose 60 – 70 mg/dL, or 71 – 85 mg/dL and decreasing: stop insulin infusion, obtain blood glucose level every 30 minutes until blood glucose > 85 mg/dL for three consecutive measurements, then check blood glucose every hour.
4. Resuming insulin infusion:

   • Restart at half the previous rate when blood glucose rises above 150 mg/dL.

Table 1:

Insulin Infusion Adjustment (Do not adjust insulin rate every hour - only make adjustments to the insulin rate every two hours)

Blood Glucose	If BG DECREASES≥ 30 mg/dl since last level	If BG isSTABLE (change in BG < 30 mg/dl) since last level	If BG INCREASES ≥ 30 mg/dl since last level
≤ 60	Stop insulin infusion See Hypoglycemia Protocol	Stop insulin infusion See Hypoglycemia Protocol	––
61 – 70	Stop insulin infusion See Hypoglycemia Protocol	Stop insulin infusion/ See Hypoglycemia Protocol	––
71–85	Stop insulin infusion See Hypoglycemia Protocol	Decrease rate by 50%
86 – 100	Decrease rate by 50%	Decrease rate by 50%	––
101 – 115	Decrease rate by 50%	Continue current rate	––
116 – 150	Decrease rate by 50%	Increase rate by 25%	Increase rate by 25%
151 – 200	Decrease rate by 25%	Increase rate by 25%	Bolus 2 units/ Increase rate by 25%
201 – 250	Continue current rate	Bolus 2 units/ Increase rate by 25%	Bolus 4 units/ Increase rate by 25%
251 – 300	Continue current rate	Bolus 4 units/ Increase rate by 50%	Bolus 6 units/ Increase rate by 50%
301 – 350	Continue current rate	Bolus 6 units/ Increase rate by 50%	Bolus 8 units/ Increase rate by 50%
351 – 400	Continue current rate	Bolus 8 units/ Increase rate by 50%	Bolus 10 units/ Increase rate by 50%
> 400	Notify staff anesthesiologist*	Notify staff anesthesiologist*	Notify staff anesthesiologist*

(Note: if insulin rate is ≥ 30 units/hr* notify staff anesthesiologist)

^* Severe hyperglycemia will be treated per anesthesiologist’s discretion.

Appendix 2. Hyperinsulinemic Normoglycemia versus Conventional Insulin Infusion for Intra-operative Glucose Management in Patients having Cardiac Surgery: A Randomized Clinical Trial

Summary of Major protocol changes from original protocol (date July 23, 2006)

Date	Protocol Change	Rationale
07/23/2006	• Original protocol
11/19/2007	• Inclusion criteria were broadened. Initial inclusion criteria were changed from patients having mitral valve surgery with CABG to all cardiac surgeries requiring cardiopulmonary bypass	An increase in patient enrollment was needed.
05/07/2008 (delirium) 07/17/2008 (echocardiographic and left atrial tissue analysis) 4/28/2008 (quality of life - deleted DASI) 05/07/2008 (quality of life, add SF-12)	The following secondary outcomes were limited to a single center sub-investigations (Cleveland Clinic only) and are thus not reported in this manuscript • Postoperative delirium • Echocardiographic-measurement of left ventricular function • Postoperative quality of life measured by a health survey • Analysis of left atrial tissue	These sub-investigations answered site-specific questions.
05/07/2008	Changed measurement of follow-up of endpoints, including all-cause mortality, from 15 – 30 days to 1 and 3 months (30 and 90 days).	This change allowed better synchronization and efficiency of data collection with other timepoints. All 15-day outcomes continued to be captured at 30 days.
9/29/2008 (hospital readmission) 01/29/2009 (additional secondary outcomes)	We revised the secondary outcomes including a “composite of minor outcomes” which including prolonged intubation, low cardiac index, renal insufficiency, prolonged hospitalization, hospital readmission	This revised secondary outcome captured perioperative data that had a lesser, but still important, impact on postoperative course.
01/29/2009	The following exclusion criteria were added: • Active infection including patients with endocarditis or infected pacemaker leads • Any infection requiring long-term antibiotics (>14 days) • Kidney disease requiring renal replacement therapy	To identify patients who required renal replacement therapy and serious infection, which are components of the primary composite outcome, we excluded patients who required renal replacement therapy or had severe infection at time of enrollment.
01/29/2009	The following primary outcomes were deleted: 1) Low cardiac index 2) Perioperative myocardial infarction 3) Prolonged intubation (>72 hours) 4) Anuria (urine output < 0.5 cc/kg/8 hr)	Changes to the primary outcome were made to improve measures of postoperative recovery. These outcomes were problematic because of 1) lack of standardized definition for perioperative myocardial infarction 2) inadequate documentation of low cardiac index and anuria 3) heterogeneity of patients who experience prolonged intubation
01/29/2009	The following secondary outcomes were removed: dysrhythmias, hyperlactatemia, postoperative troponin I or T, inotropic support	Secondary outcomes were revised to select more important clinical endpoints
01/29/2009	The sample size was increased to 2790 patients and the number of interim analyses was changed from 3 to 7	The sample size estimate was re-adjusted to power the study for a 30% reduction in risk of complications (previously a 60% reduction).

Appendix 3.

Appendix 3A.

Primary outcome: Definitions of the components of a composite of major post-operative complications occurring within 30 days after surgery

Major complications	Requirements for acceptance
Death within 30 days	All-cause mortality identified during initial hospitalization or during 30 day follow-up.
Post-operative mechanical circulatory support	Failure to wean from CPB or post-operative low cardiac index (CI <1.8 l/min/m2) conditions requiring circulatory support with either intra-aortic balloon pump, ventricular assist device, and/or extracorporeal mechanical oxygenation during or post-CPB or during post-operative course.
Serious Infection Morbidity	Post-operative course complicated by one of the following: 1) sepsis with evidence of acute organ dysfunction. Sepsis is recognized as a clinical syndrome that may be defined by infection highly suspected (clinical syndrome pathognomonic for infection) or proven (by culture, stain, or polymerase chain reaction) and presence of two or more of the following systemic inflammatory response syndrome (SIRS) criteria: heart rate > 90 beats per minute (tachycardia); body temperature < 36 °C or > 38 °C (hypothermia or fever); respiratory rate > 20 breaths per minute or, a PaCO2 less than 32 mmHg (tachypnea or hypocapnia due to hyperventilation); white blood cell count < 4000 cells/mm3 or > 12000 cells/mm3, or greater than 10% band forms (immature white blood cells) (leukopenia, leukocytosis, or bandemia)124; 2) mediastinitis (sternal click, open sternal wound, drainage from mediastinal incision, with fever, and including positive cultures along with elevated white blood cell count and the institution of antimicrobial therapy and reexploration with operative note diagnosing mediastinitis or sternectomy with muscle flap grafts to the affected area or diagnosis by physician of mediastinitis); 3) sternal wound infection (sternal wound infection other than mediastinitis, documented with positive cultures, requiring surgical intervention); or 4) pneumonia (fever > 38°C, elevation in white blood cell count, increase in sputum production, infiltrate in chest x-ray > 24 hours, positive sputum culture) requiring mechanical ventilation.
Renal morbidity	Post-operative requirement for renal dialysis. Patients with pre-operative requirement for dialysis are excluded.
Neurologic deficit	New post-operative focal (aphasia, decrease in limb function, or hemiparesis confirmed by clinical findings and/or computed tomographic scan) or global neurologic deficit (diffuse encephalopathy with greater than 24 hours of severely altered mental status, and/or failure to awaken post-operatively).

Appendix 3B.

Secondary Outcomes

Secondary Outcomes	Requirements for Acceptance
Composite of minor outcomes	The occurrence of one of more of the minor complications listed in Table 3 occurring within 30 days of surgery
Postoperative atrial fibrillation	The occurrence of new-onset postoperative atrial fibrillation following cardiac surgery occurring within 30 days of surgery. Patients who had paroxysmal or persistent atrial fibrillation prior to surgery are excluded.
Duration of Hospitalization	Days from day of surgery to hospital discharge.
Duration of Intensive Care Unit Stay	Days from day of surgery to discharge from intensive care unit.
All-cause mortality at one year	All-cause mortality identified during one-year follow-up.

Appendix 3C.

Components of the composite of the minor outcomes (a secondary outcome).

Prolonged intubation	Endotracheal intubation and mechanical ventilation required for >72 hours post-operatively, measured from arrival in intensive care unit following surgery until weaning from mechanical ventilation and endotracheal extubation. Additional periods of time where reintubation and mechanical ventilation is required are included.
Low cardiac index	Cardiac index <1.8 l/min/m2 despite adequate fluid replacement and high dose inotropic support for >4 hours
Renal insufficiency	Postoperative increase in baseline creatinine >100%. Baseline creatinine was defined as the preoperative measurement immediately prior to surgery.
Prolonged hospitalization	Hospitalization following surgery > 30 days
Hospital readmission	Postoperative complications requiring readmission to a hospital for any reason identified during 30 day followup.

Appendix 4. Treatment effect on secondary composite outcome by site

Site Complications	Hyperinsulinemic Normoglycemia	Standard Therapy	Relative Risk (99.92% CI)*	P-value
Cleveland Clinic, Cleveland	N = 252	N = 266
Postoperative atrial fibrillation	108 (44)	128 (50)	0.89 (0.64,1.23)	0.23
Duration of hospitalization (days)	7 [5, 12]	7 [5, 11]	0.93 (0.69, 1.25)*	0.40
Intensive care unit stay (hours)	41(27.8,46.2)	32 (27.7, 5.2)	0.99 (0.74, 1.33)*	0.89
One-year all-cause mortality	18 (7)	8 (3)	2.38 (0.59, 9.52)	0.031
Any minor complication c	119 (47)	127 (48)	0.99 (0.72, 1.36)	0.89
Royal Victoria Hospital, Montreal	N = 457	N = 464
Postoperative atrial
fibrillation	101(22)	107(23)	0.96 (0.64, 1.44)	0.73
Duration of hospitalization(days)	8 [7, 11]	9 [7, 13]	1.13 (0.90, 1.42)*	0.066
Intensive care unitstay (hours)	24(22.8,24.5)	24(23.6,25.0)	1.24 (0.98, 1.57)*	0.0026
One-year all-cause mortality	14 (3)	14 (3)	1.04 (0.30, 3.59)	0.92
Any minor complication c	81 (18)	110 (24)	0.77 (0.49, 1.21)	0.051

Data are presented as median [25^th, 75^th percentiles] for length of ICU stay and hospital stay, event (%) for binary outcomes.

^* Hazard ratios

^aConfidence intervals adjusted for group sequential design using confidence coefficient of 2.633 for combined sites and 2.86 within sites in order to maintain overall study alpha of 0.05.

^b Chi-square test for binary outcomes, and Cox proportional hazards model for length of ICU stay and hospital stay. Bonferroni correction: significant if P < 0.0042/5 = 0.00084 within site.

^cincluded mechanical ventilation >72 hours, low cardiac index, acute kidney injury, hospitalization >30 days, all-cause hospital readmission within 30 days.