Clinical outcomes of hybrid coronary revascularization versus coronary artery bypass surgery in patients with diabetes mellitus Ralf E. Harskamp, MD, a,b Patrick F. Walker, BS, c John H. Alexander, MD, MHS, a Ying Xian, MD, PhD, a Henry A. Liberman, MD, c Robbert J. de Winter, MD, PhD, b Thomas A. Vassiliades, MD, MBA, c Eric D. Peterson, MD, MPH, a John D. Puskas, MD, MS, c and Michael E. Halkos, MD, MS c Durham, NC; The Netherlands and Atlanta, GA

Background Hybrid coronary revascularization (HCR) involves minimally invasive left internal mammary artery to left anterior descending coronary artery grafting combined with percutaneous coronary intervention (PCI) of non–left anterior descending vessels. The safety and efficacy of HCR among diabetic patients are unknown. Methods Patients with diabetes were included who underwent HCR at a US academic center between October 2003 and September 2013. These patients were matched 1:5 to similar patients treated with coronary artery bypass grafting (CABG) using a propensity score (PS)-matching algorithm. Conditional logistic regression and Cox regression stratified on matched pairs were performed to evaluate the association between HCR and inhospital complications, a composite measure of 30-day mortality, myocardial infarction and stroke, and up to 3-year all-cause mortality. Results

Of 618 patients (HCR = 103; CABG = 515) in the PS-matched cohort, the 30-day composite of death, MI, or stroke after HCR and CABG was 4.9% and 3.9% (odds ratio: 1.25; 95% CI [0.47-3.33]; P = .66). Compared with CABG, HCR also had similar need for reoperation (7.6% versus 6.3%; P = .60) and renal failure (4.2% versus 4.9%; P = .76) but required less blood products (31.4% versus. 65.8%; P b .0001), lower chest tube drainage (655 mL [412-916] versus 898 mL [664-1240]; P b .0001), and shorter length of stay (b5 days: 48.3% versus 25.3%; P b .0001). Over a 3-year follow-up period, mortality was similar after HCR and CABG (12.3% versus 14.9%, hazard ratio: 0.94, 95% CI [0.47-1.88]; P = .86).

Conclusion Among diabetic patients, the use of HCR appears to be safe and has similar longitudinal outcomes but is associated with less blood product usage and faster recovery than conventional CABG surgery. (Am Heart J 2014;168:471-8.)

Coronary artery disease (CAD) is a major cause of morbidity and mortality among patients with diabetes mellitus. 1 Compared with nondiabetic patients, those with diabetes also have substantially worse outcomes. 1,2 Findings from recent studies demonstrate improved survival with coronary artery bypass grafting (CABG) surgery compared with percutaneous coronary intervention (PCI) with drug-eluting stents (DES) in diabetic patients with multivessel CAD. 3,4 The

From the aDuke Clinical Research Institute, Durham, NC, bAcademic Medical Center– University of Amsterdam, Amsterdam, The Netherlands, and cDivisions of Cardiothoracic Surgery and Cardiology, Clinical Research Unit, Emory University School of Medicine, Atlanta, GA. Submitted April 16, 2014; accepted June 30, 2014. Reprint requests: Ralf E Harskamp, MD, 2400 Pratt St, Room 7047B, Durham NC 27705. E-mail: [email protected] 0002-8703 © 2014, Mosby, Inc. All rights reserved. http://dx.doi.org/10.1016/j.ahj.2014.06.027

observed survival benefit of CABG in these patients is presumed to be largely attributable to the left internal mammary artery (LIMA) to left anterior descending coronary artery (LAD) bypass, which provides excellent long-term durability. 5 However, CABG is also associated with significant inhospital morbidity, including stroke. In addition, the superiority of CABG over PCI in PCI amenable non-LAD lesions is questionable, given saphenous vein graft failure rates of approximately 10% to 20% within 1 year and ≥50% at 10 years. 6 7 Therefore, hybrid coronary revascularization (HCR) has been proposed as a combined surgical and percutaneous approach, which combines the benefits of a LIMA-LAD graft with the use of coronary stents for non-LAD lesions. In addition, when LIMA-to-LAD grafting is performed through limited incisions, HCR may also present a less invasive alternative to conventional CABG. A number of studies have been performed that suggest that HCR may result in faster recovery, fewer complications, and equivalent clinical outcomes compared

American Heart Journal October 2014

472 Harskamp et al

with CABG in a general population of patients with multivessel CAD. 8-11 The purpose of this study was to compare clinical outcomes after HCR versus conventional CABG surgery in a matched high-risk population of patients with diabetes mellitus. We specifically sought to compare 30-day and longitudinal clinical outcomes, inhospital complications, and recovery.

Figure 1

Methods Study population and definitions The starting population for the current analysis all eligible cases included in the Emory University institutional Society of Thoracic Surgeons (STS) Adult Cardiac Database (www.STS.org) between October 2003 and September 2013. At Emory, a custom data field was created within the STS database that defined hybrid patients on an intent-to-treat basis, in which HCR procedures involved a planned nonsternal LIMA-LAD bypass with PCI of ≥1 non-LAD lesion that were performed either in one setting or as 2-staged procedures. From this starting population of HCR “intention to treat” and other cardiac surgery procedures, we only selected patients with a history of diabetes (n = 4,032) (both insulin and noninsulin requiring) and applied a number of exclusion criteria, which are listed in Figure 1. From the remaining 3,427 patients, we then matched HCR cases (“as treated”) to control patients who underwent elective or urgent conventional on- or off-pump CABG surgery. We also performed a sensitivity analysis in which diabetic patients were compared in whom HCR was considered (intention to treat) with a matched cohort of patients who underwent conventional CABG surgery. Diagnostic criteria for all characteristics other than HCR, including diabetes, were based on the STS registry definitions. This study was conducted in accordance with the Institutional Review Board approval of Emory University. The authors are solely responsible for the design and conduct of this study, all study analysis, the drafting and editing of the paper, and its final contents. No extramural funding was used to support this work. Indications, contraindications, and procedural information At Emory, the consideration for HCR as well as the timing and sequence of the surgical and percutaneous components are carefully discussed with the heart team, but also with referring cardiologists as well as with the patient. The indications and contraindications for a hybrid approach as well as details on procedural information on HCR at Emory have been published previously. 12,13 In short, relative angiographic indications for HCR were presence of a significant stenosis in the proximal LAD disease that is amenable to LIMA-LAD bypass and non-LAD lesions that have an anatomy that is amenable to PCI. Relative contraindications for HCR included hemodynamic instability, prior cardiac or

Title: Flow diagram of the study population.Legend: Shown are the steps that led from the starting population to the PS-matched study population. ⁎These patients were included in a sensitivity analysis of intention to treat for HCR.

thoracic surgery, severe lung disease with the inability to tolerate single-lung ventilation, and morbid obesity. In most HCR cases, LIMA-to-LAD revascularization is performed first, with administration of clopidogrel (150 mg) approximately 4 hours after LIMA-LAD bypass and an additional loading dose of 300 mg at the time of PCI. This approach minimizes the risks of bleeding complications due to anticoagulation and dual-antiplatelet therapy use and allows assessment of patency of the LIMA-LAD graft at the time of PCI. However, patients with critical non-LAD anatomy PCI is usually performed first. In these cases, LIMA-LAD grafting was performed while continuing clopidogrel (75 mg/daily). For patients who underwent one-stage HCR procedures, 600 mg clopidogrel was administered through nasogastric tube after confirmation of LIMA patency but before coronary stenting. Most patients underwent both procedures during the index hospitalization to ensure complete revascularization. At Emory, the surgical component of HCR was performed with an endoscopic atraumatic coronary artery bypass approach up to 2009; thereafter, LIMA harvest was performed with the use of robotic assistance (Da Vinci Robotic surgical system; Intuitive Surgical, Sunnyvale, CA). After identification of the optimal target site on the LAD, the LIMA-to-LAD anastomosis is subsequently performed through a nonrib spreading, minithoracotomy, using a minimally invasive stabilizer (Nuvo; Medtronic, Minneapolis, MN). The whole procedure is performed without the use of cardiopulmonary bypass. The PCI component of HCR was performed using standardized methods and techniques. In most cases, coronary stent placement involved either first (sirolimus and placlitaxel) or newer (everolimus and zotarolimus) generation DES.

American Heart Journal Volume 168, Number 4

Harskamp et al 473

Table I. Baseline characteristics before and after PS matching Unadjusted

Age Female White BMI Current smoker Insulin use Chron. lung dis. Dyslipidemia Hypertension Cerebrovasc. dis. Periph. art. dis. Renal failure Prior MI LVEF, % 3-vessel dis Left main N50% Hemoglobin Hg A1c (%) Urgent Preoperative medications β-blockers⁎ Salicylate use⁎ ADP inhibitor use b5 d⁎

After PS matching

HCR

CABG

Stand

HCR

CABG

Stand

(n = 103)

(n = 3,307)

Diff

(n = 103)

(n = 515)

Diff

65.2 ± 10.6 36 (35.0) 73 (70.9) 29.5 ± 5.4 21 (20.4) 36 (35.0) 5 (4.9) 98 (95.1) 99 (96.1) 23 (22.3) 13 (12.6) 12 (11.7) 48 (46.6) 55.0 ± 9.8 59 (57.3) 20 (19.4) 12.6 ± 1.8 7.2 ± 1.4 43 (41.7)

63.2 ± 10.2 1073 (32.4) 2377 (71.9) 31.4 ± 14.0 765 (23.1) 1180 (35.7) 210 (6.4) 2927 (88.5) 3093 (93.5) 704 (21.3) 635 (19.2) 343 (10.4) 1832 (55.4) 50.7 ± 12.8 2537 (76.7) 908 (27.5) 12.7 ± 2.3 7.7 ± 2.0 1568 (47.4)

19.6 5.2 2.2 35.0 6.8 1.5 6.9 30.7 13.3 2.5 19.7 4.0 17.5 43.6 39.1 20.2 4.8 37.4 11.4

65.2 ± 10.6 36 (35.0) 73 (70.9) 29.5 ± 5.4 21 (20.4) 36 (35.0) 5 (4.9) 98 (95.1) 99 (96.1) 23 (22.3) 13 (12.6) 12 (11.7) 48 (46.6) 55.0 ± 9.8 59 (57.3) 20 (19.4) 12.6 ± 1.8 7.2 ± 1.4 43 (41.7)

64.8 ± 9.9 174 (33.8) 367 (71.3) 29.5 ± 5.8 109 (21.2) 177 (34.4) 24 (4.7) 493 (95.7) 491 (95.3) 114 (22.1) 69 (13.4) 57 (11.1) 243 (47.2) 54.9 ± 11.0 296 (57.5) 90 (17.5) 12.8 ± 3.8 7.3 ± 1.5 226 (43.9)

4.6 2.4 0.9 1.1 1.9 1.2 0.9 2.7 4.0 0.5 2.3 1.8 1.2 0.9 0.4 4.9 9.3 4.0 4.3

90 (87.4) 87 (84.5) 19/75 (25.3)

3117 (79.8) 3137 (80.3) 254/2013 (12.6)

31.3 10.4 19.0

90 (87.4) 87 (84.5) 19/75 (25.3)

408 (79.2) 418 (81.2) 34/250 (12.0)

21.5 7.9 33.7

Abbreviations: Stand Diff, standardized difference; BMI, Body mass index; Chron. lung dis., chronic lung disease; Cerebrovasc. dis., cerebrovascular disease; Periph. art. dis., peripheral artery disease; LVEF, left ventricular ejection fraction; 3-vessel dis, 3-vessel disease; Hg A1C, glycated hemoglobin; ADP, adenosine diphosphate receptor inhibitor. Data are presented as number and percentage or mean and SD. ⁎ Preoperative medications were not included in the propensity-matched models.

Outcomes Outcomes of interest included the composite of death, stroke, and myocardial infarction (MI) at 30 days as well as a number of inhospital outcomes, including any reoperation, renal failure, prolonged ventilation (N24 hours), deep sternal wound infection, CABG-related bleeding, excessive use of blood transfusions (≥5 U) or chest tube drainage (≥2 L within 24 hours after surgery), and short and long postoperative length of stay. Definitions for these outcomes are based on STS Adult Cardiac Surgery Database definitions. In addition, we compared all-cause mortality up to 3 years between the 2 groups, in which follow-up information was obtained by querying the Social Security Death Index to determine dates of death for all patients in the study. Statistical analysis After applying the exclusion criteria as listed in Figure 1, we performed PS matching to reduce the effects of treatment selection bias. The PS (or the probability of assignment to HCR or CABG) was constructed using multivariable logistic regression from available demographics, clinical, and angiographic characteristics, except

for preoperative medications (see Table I). Patients who underwent HCR were matched in a 5:1 ratio with patients who underwent CABG, using the nearest neighbor-matching algorithm without replacement on the logit of the PS using a caliper of width equal to 0.20 SDs of the logit of PS. 14 After PS matching, we tested whether the balance on the covariates was achieved through the matching procedures using standardized differences (an absolute standardized difference of b10% was considered a negligible imbalance 15) and assessment of global imbalance (P = .99 after PS matching). 16 Comparisons for inhospital outcomes and 30-day composite end point were performed using conditional logistic regression analyses accounting for the matched nature of the PS-matched sample. The Kaplan-Meier method was used to create survival curves, and Cox proportional hazard regression analysis stratified on matched pairs were used to examine differences in mortality up to 3 years after the procedure. The proportional hazards assumptions were evaluated by visual assessment by log minus log survival plots and were found valid. Statistical analyses were performed with IBM SPSS Statistics version 20.0 (IBM Corporation, Armonk, NY). Propensity score matching was performed using an SPSS-R plugin 17 for R packages (MatchIt, Ritools, and cem).

American Heart Journal October 2014

474 Harskamp et al

Results From 2003 to 2013, 118 patients with diabetes underwent HCR on an intent-to-treat basis. Of those who were scheduled for HCR, 15 patients (12.7%) did not undergo one or both stages of HCR. Most (n = 12) of these patients underwent minimally invasive LIMA-LAD, but the other lesions were medically managed instead of treated with PCI. In 2 patients, CABG was performed instead of HCR, and one patient died before HCR could be performed. The remaining 103 cases all underwent HCR, either as a one-stage procedure (n = 4; 3.9%), or a staged procedure, in which PCI was performed first in 19 cases (18.4%), and CABG was performed first in 80 cases (77.7%). Most HCR cases were performed during the same hospitalization (n = 84 or 81.6%). The median time interval between the 2 stages was 3 days (25 th-75 th percentile: 2-7 days).

Clinical and procedural characteristics A total of 103 diabetic patients who underwent HCR with minimally invasive LIMA-LAD bypass and PCI of non-LAD coronary vessels were matched 1:5 with conventional CABG patients based on 19 preoperative variables. As shown in Table I, demographics and clinical characteristics were well balanced after PS matching (standardized difference b10%). In the matched cohort, a minority of patients in both groups required insulin to adequately manage blood glucose levels, namely, 35.0% of those who underwent HCR and 34.4% of those scheduled for CABG. All other patients were on oral agent treatment with or without diet treatment. As shown in Table II, total operating time tended to be shorter after HCR compared with CABG, and the use of cardiopulmonary bypass as well as bilateral internal mammary artery (IMA) use was higher in patients undergoing CABG. In the HCR group, PCI was performed in ≥1 other non-LAD vessel for complete revascularization, in which the vast majority of patients received a drug-eluting coronary stent (90.3%). Short-term outcomes The primary objective of the study was to examine the incidence of the composite of death, MI, and stroke at 30 days. As shown in Table III, the 30-day composite outcome was similar between HCR and CABG (4.9% versus 3.9%; P = .66). There were 2 deaths in the HCR group versus 10 deaths in the CABG group (1.9% versus 2.1%,), 1 versus 4 patients had MIs (1.0% versus 0.8%), and 2 versus 9 had strokes (1.9% versus 1.7%). Although inhospital complications, such as renal failure, need for reoperation, and prolonged ventilation, were similar between the 2 groups, the use of blood products, chest tube drainage were fewer, and postoperative recovery was faster after HCR, as shown by shorter postoperative hospital length of stay.

Table II. Procedural characteristics Characteristic

HCR (n = 103)

Total operating time, min 284 (284-329) (median, IQR) Cardiopulmonary bypass use (%) 1 (1.0) Bilateral IMA use (%) 1 (1.0) IMA anastomosis (median, IQR) 1 (1, 1) Vein grafts (median, IQR) 0 (0, 0) Total grafts (median, IQR) 1 (1,1) No. of stents used (median, IQR) 2 (1, 2) Stent use Balloon angioplasty 2 (1.9) Bare metal stent 8 (7.8) DES 92 (89.3) Drug-eluting and 1 (1.0) bare-metal stent PCI treated vessels RCA or side branch 27 (31.4) CX or side branch 24 (27.9) CX and RCA 9 (10.5) LM 9 (10.5) Diagonal 5 (5.8) LM and CX 4 (4.6) LM and RCA 4 (4.6) Diagonal and CX 2 (2.3) Ramus intermedius 2 (2.3)

CABG (n = 515)

Stand diff

312 (262, 363) 9.4 83 (16.1) 47 (9.1) 1 (1, 1) 2 (1, 3) 3 (3, 4) NA NA

55.7 37.9 32.3 248.4 278.5 NA NA

NA

NA

Abbreviations: IQR, Interquartile range; NA, not applicable; LM, left main coronary artery; CX, circumflex coronary artery; RCA, right coronary artery.

Longitudinal clinical outcomes The secondary end point of this study was to assess and compare midterm clinical outcomes. The median follow-up time was 2.9 years (25 th-75 th percentiles: 1.4-3.0). As illustrated in Figure 2, mortality was similar after HCR and CABG over 3 years of follow-up (hazard ratio [HR]: 0.94 [0.47-1.88]; P = .86). Sensitivity analysis A sensitivity analysis was performed in which patients with HCR as an intention to treat (HCR-ITT) (n = 118) were matched to patients who underwent conventional CABG (n = 590). As shown in Table IV, the composite of 30-day death, MI, and stroke was similar between these 2 groups. Other inhospital outcomes showed less frequent use of blood products (31% versus 66%) and shorter postoperative length of stay in the HCR-ITT group. Three-year mortality estimates were comparable for the HCR-ITT group and the matched cohort of CABG patients (16.3% versus 15.2%; overall HR: 1.29 [0.72-2.32], P = .38).

Discussion Over the past decade, minimally invasive cardiac surgery has grown in popularity for the treatment of single- and multivessel CAD, in a desire to further optimize outcomes after surgery. 18 In selected patients

American Heart Journal Volume 168, Number 4

Harskamp et al 475

Table III. Thirty-day major adverse cerebrovascular and cardiovascular events and inhospital outcomes

Characteristic Composite of 30-day death, MI, stroke Death MI Stroke Inhospital major morbidity Reoperation Renal failure Prolonged ventilation (N24 h) Access site infection Other outcomes CABG-related bleeding Use of blood products Chest tube drainage (mL/24 h) Hypoglycemia (≤70 mg/dL) Short PLOS (b5 d) Long PLOS (N14 d)

HCR

CABG

(n = 103)

(n = 515)

OR (95% CI)

5 2 1 2

(4.9) (1.9) (1.0) (1.9)

20 10 4 9

(3.9) (1.9) (0.8) (1.7)

1.25 1.00 1.25 1.11

7 4 8 0

(6.8) (3.9) (7.8) (0.0)

25 24 55 10

(4.9) (4.7) (10.7) (1.9)

1.40 (0.61-3.24) 0.83 (0.29-2.40) 0.73 (0.35-1.53) –

.43 .74 .40 –

(11.5) (59.8) (700-1250) (2.1) (29.1) (7.2)

0.75 (0.37-1.51) 0.49 (0.34-0.71) ⁎

.42 b.0001 b.0001 .45 .003 .24

9 30 650 1 49 4

(8.7) (29.1) (420-855) (1.0) (47.6) (3.9)

59 308 935 11 150 37

(0.47-3.33) (0.22-4.56) (0.14-11.18) (0.24-5.14)

P

0.46 (0.06-3.52) 1.63 (1.18-2.25) 0.54 (0.19-1.52)

.66 1.00 .84 .89

Abbreviations: OR, Odds ratio; CI, confidence interval; PLOS, postoperative length of stay. Data are presented as number and percentage between parentheses. ⁎ β coefficient: −2.17, t = −5.51; P b .0001.

Figure 2

Title: Kaplan-Meier estimates of all-cause mortality.Legend: Shown are rates of death from any cause truncated at 3 years after surgery. The HR was calculated using a Cox modeling stratified on matched pairs. Abbreviations: HCR, hybrid coronary revascularization.

with multivessel CAD, hybrid revascularization provides a viable alternative to both conventional CABG and multivessel PCI. 8-11 To our knowledge, this study represents the first experience of HCR to conventional CABG in diabetic patients. We found that, among patients with

diabetes mellitus, HCR (either performed “as treated” or on an “intention-to-treat” basis) is associated with similar short-term mortality and morbidity as well as longitudinal outcomes compared with conventional CABG surgery. In contrast, HCR was associated with lower blood utilization and significantly faster postoperative recovery than conventional surgery.

Surgery and coronary stenting in diabetic patients with multivessel disease Several randomized studies have compared surgical versus percutaneous revascularization among diabetic patients with multivessel CAD. However, these studies mostly involved substudies of larger randomized clinical trials. 19-21 A landmark study was the Future Revascularization Evaluation in Patients With Diabetes Mellitus: Optimal Management of Multivessel Disease (FREEDOM) trial, which enrolled 1,900 diabetic patients who were 1:1 randomized to CABG or PCI with DES. 4 The main findings of this study included a 30% reduction in the composite of 5-year death, MI, and stroke (18.7% versus 26.6%; P = .005) after CABG when compared with PCI, which was driven by a reduction in all-cause mortality (6.3% versus 10.9%; P = .049) and nonfatal MI (6.0% versus 13.9%; P b .001), whereas stroke rates were significantly higher in the surgical group (5.2% versus 2.4%; P = .03). A recent metaanalysis summarized the data from 8 randomized trials and found that at a mean or median 5-year follow-up, diabetic patients (n = 3,131) allocated to CABG had lower all-cause mortality compared with those allocated to PCI (risk ratio: 0.67; 95% CI [0.52-0.86]; P = .002). 3 No difference in mortality was observed when PCI was performed with DES

American Heart Journal October 2014

476 Harskamp et al

Table IV. Sensitivity analysis: 30-day major adverse cerebro- and cardiovascular events and in-hospital outcomes for hybrid intent-to-treat versus CABG cases

Characteristic Composite of 30-day death, MI, stroke Death MI Stroke Inhospital major morbidity Reoperation Renal failure Prolonged ventilation (N24 h) Sternal wound infection Other outcomes CABG-related bleeding Use of blood products Chest tube drainage (mL/24 h) Hypoglycemia (≤70 mg/dL) Short PLOS (b5 d) Long PLOS (N14 d)

HCR-ITT

CABG

(n = 118)

(n = 590)

5 3 1 2

OR (95% CI)

P

(4.2) (2.5) (0.8) (1.7)

19 7 3 10

(3.2) (1.2) (0.5) (1.7)

1.32 2.14 1.67 1.00

(0.49-3.52) (0.55-8.29) (0.17-16.02) (0.22-4.56)

.59 .27 .66 1.00

9 (7.6) 5 (4.2) 10 (8.4) 0 (0.0)

37 29 90 10

(6.3) (4.9) (15.3) (1.7)

1.22 0.86 0.56 0.04

(0.59-2.52) (0.33-2.23) (0.29-1.07) (0.00-50.46)

.60 .76 .078 .37

(13.7) (65.8) (664-1240) (2.9) (25.3) (7.6)

0.68 (0.36-1.28) 0.48 (0.34-0.67) ⁎

11 (9.3) 37 (31.4) 655 (412-916) 1 (0.8) 57 (48.3) 6 (5.1)

81 388 898 17 149 45

0.39 (0.04-2.21) 1.91 (1.41-2.60) 0.67 (0.28-1.56)

.23 b.0001 b.0001 .23 b.0001 .35

Data are presented as number and percentage between parentheses. ⁎ β coefficient: −1.91, t = −5.12; P b .0001.

or bare-metal stents. Other studies that compared various stent designs among diabetic patients did show that second-generation DES outperform first-generation and bare-metal stents, both on angiographic as well as nonfatal clinical end points, including late restenosis and need for repeat revascularization. 2,22

A role for HCR in diabetic populations The use of HCR may provide an attractive alternative to both multivessel PCI and CABG in diabetic patients. The use of the LIMA-to-LAD graft has been postulated to play a key role in the improved survival with CABG over PCI in diabetic patients. 4,5 As such, based on the routine use of the LIMA to bypass the LAD for treatment of proximal LAD disease, one would expect that survival rates after HCR would be equivalent to those seen after CABG. Based on follow-up data that were limited to 3 years of follow-up, we did indeed observe similar midterm clinical outcomes after HCR and CABG. A major drawback for conventional CABG surgery is the risk of procedure-related stroke. In the FREEDOM trial, a 6-time excess of 30-day stroke was seen in the CABG group compared with multivessel PCI (1.8% versus 0.3%; P = .002). It has been postulated that the avoidance of cardiopulmonary bypass as well as aortic cross/side clamping in HCR may result to less procedurerelated strokes after compared with CABG and hereby improve short-term outcomes. 11,23 Prior studies have indeed shown low rates of stroke (b0.5%) when minimal invasive surgical techniques were used. 24,25 In our study, the rates of stroke were equivalent after both procedures (1.9% and 1.7%) and were also very similar to these seen in the CABG group of the FREEDOM trial. However, the less invasive nature of HCR did translate to

shorter postprocedural hospital length of stay. Although reasons for this difference are based on speculation, we did see lower blood product use and chest tube drainage and numerical differences in ventilation time, renal failure, and wound infections.

Future directions Although our study suggests that HCR can be safely performed in diabetic patients, these findings will have to be further explored in other centers, to assess the consistency of results across various centers, using various HCR approaches and techniques. Careful patient selection for HCR is important to optimize long-term clinical outcomes. As shown in prior studies, diabetic patients with lesions that are complex, including multisegment and diffuse disease, do not fare well, particularly after PCI. As such, when these lesions are located in nonLAD territories, these patients may not provide ideal candidates for HCR but may be better candidates for conventional CABG surgery. Recently, a first attempt toward a structural approach for selecting patients for HCR has been proposed, in which the heart team will play a central role in decision making. 26 Future studies involving HCR that use a CAD severity matrix (such as the SYNTAX score) may provide better guidelines and selection criteria for diabetic patients undergoing HCR. Studies that assess predictors for poor outcomes after HCR or validate existing surgical risk scores are also warranted, as they can help in preoperative assessment of potential candidates for HCR and ultimately optimize outcomes. Finally, future randomized clinical studies will have to be designed and performed that compare clinical outcomes after conventional coronary revascularization.

American Heart Journal Volume 168, Number 4

Limitations Our study had several limitations. First, this study involves a nonrandomized comparison of 2 treatment strategies, which by definition is prone to selection bias. Although we used PS matching on a number of identified confounders, the issue of unmeasured confounding remains. Second, confounding due to competing risk (patients who die while being considered for CABG or HCR as well as patients who die between the 2 stages of HCR) remains, although we tried to account for this by performing a sensitivity analysis based on an intentionto-treat versus as-treat analysis. Third, due to limitations of the database, detailed angiographic information, such as SYNTAX score, vessel size, target vessel quality, and the extent of diffuse distal CAD, was not collected. Finally, although information on midterm survival was collected, death from cardiovascular causes as well as information on additional outcomes, such as MI and stroke was not collected.

Conclusion Among patients with diabetes, the use of HCR appears to be safe, with faster postoperative recovery and similar midterm clinical outcomes when compared with conventional coronary bypass surgery. Definite conclusions regarding safety and efficacy of HCR would need to be addressed in randomized trials.

Disclosures Dr Alexander receives research grants from Bristol Myers Squibb, CLS Behring, and Regado Biosciences. He also serves as a consultant for Moerae Matrix and VA Cooperative Studies program. Dr Vassiliades is the vice president medical affairs, structural heart at Medtronic. Dr Peterson is a consultant for Boehringer Ingelheim, Genentech, Janssen Pharmaceutical Products, Merck, and Sanofi-Aventis and receives research grants from Eli Lilly and Janssen Pharmaceutical Products. Dr Halkos serves as a consultant for Intuitive Surgical. The remaining authors do not report a conflict of interest.

References 1. Beckman JA, Paneni F, Cosentino F, et al. Diabetes and vascular disease: pathophysiology, clinical consequences, and medical therapy: part II. Eur Heart J 2013;34:2444-52. 2. Harskamp RE, Park DW. Percutaneous coronary intervention in diabetic patients: should choice of stents be influenced? Expert Rev Cardiovasc Ther 2013;11:541-53. 3. Verma S, Farkouh ME, Tanagawa B, et al. Comparison of coronary artery bypass surgery and percutaneous coronary intervention in patients with diabetes: a meta-analysis of randomised controlled trials. Lancet Diabetes Endocrinol 2013;1:317-28.

Harskamp et al 477

4. Farkouh ME, Domanski M, Sleeper LA, et al. Strategies for multivessel revascularization in patients with diabetes. N Engl J Med 2012;367:2375-84. 5. Loop FD, Lytle BW, Cosgrove DM, et al. Influence of the internal-mammary-artery graft on 10-year survival and other cardiac events. N Engl J Med 1986;314:1-6. 6. Harskamp RE, Lopes RD, Baisden CE, et al. Saphenous vein graft failure after coronary artery bypass surgery: pathophysiology, management, and future directions. Ann Surg 2013;257: 824-33. 7. Byrne JG, Leacche M, Vaughan DE, et al. Hybrid cardiovascular procedures. JACC Cardiovasc Interv 2008;1:459-68. 8. Bachinsky WB, Abdelsalam M, Boga G, et al. Comparative study of same sitting hybrid coronary artery revascularization versus off-pump coronary artery bypass in multivessel coronary artery disease. J Interv Cardiol 2012;25:460-8. 9. Halkos ME, Vassiliades TA, Douglas JS, et al. Hybrid coronary revascularization versus off-pump coronary artery bypass grafting for the treatment of multivessel coronary artery disease. Ann Thorac Surg 2011;92:1695-701. 10. Shen L, Hu S, Wang H, et al. One-stop hybrid coronary revascularization versus coronary artery bypass grafting and percutaneous coronary intervention for the treatment of multivessel coronary artery disease: 3-year follow-up results from a single institution. J Am Coll Cardiol 2013;61:2525-33. 11. Harskamp RE, Zheng Z, Alexander JH, et al. Status quo of hybrid coronary revascularization for multi-vessel coronary artery disease. Ann Thorac Surg 2013;2013:460-8. 12. Halkos ME, Walker PF, Vassiliades TA, et al. Clinical and angiographic results after hybrid coronary revascularization. Ann Thorac Surg 2014;97:484-90. 13. Vassiliades Jr TA, Douglas JS, Morris DC, et al. Integrated coronary revascularization with drug-eluting stents: immediate and seven-month outcome. J Thorac Cardiovasc Surg 2006;131: 956-62. 14. Austin PC. Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies. Pharm Stat 2011;10:150-61. 15. Austin PC, Grootendorst P, Anderson GM. A comparison of the ability of different propensity score models to balance measured variables between treated and untreated subjects: a Monte Carlo study. Stat Med 2007;26:734-53. 16. Iacus SM, King G, Porro G. CEM: coarsened exact matching software. J Stat Softw 2009;30:1-27. 17. Propensity score matching in SPSS. 2012. 18. Iribarne A, Easterwood R, Chan EY, et al. The golden age of minimally invasive cardiothoracic surgery: current and future perspectives. Future Cardiol 2011;7:333-46. 19. Harskamp RE, Park DW. Optimal choice of coronary revascularization and stent type in diabetic patients with coronary artery disease. Cardiol Ther 2013;2:69-84. 20. Roffi M, Angiolillo DJ, Kappetein AP. Current concepts on coronary revascularization in diabetic patients. Eur Heart J 2011;32: 2748-57. 21. Ellis SG. Coronary revascularization for patients with diabetes: updated data favor coronary artery bypass grafting. J Am Coll Cardiol 2013;61:817-9. 22. Bangalore S, Kumar S, Fusaro M, et al. Outcomes with various drug eluting or bare metal stents in patients with diabetes mellitus: mixed treatment comparison analysis of 22,844 patient years of follow-up from randomised trials. BMJ 2012;345: e5170.

478 Harskamp et al

23. Deb S, Wijeysundera HC, Ko DT, et al. Coronary artery bypass graft surgery vs percutaneous interventions in coronary revascularization: a systematic review. JAMA 2013;310:2086-95. 24. McGinn Jr JT, Usman S, Lapierre H, et al. Minimally invasive coronary artery bypass grafting: dual-center experience in 450 consecutive patients. Circulation 2009;120:S78-84.

American Heart Journal October 2014

25. Halkos ME, Liberman HA, Devireddy C, et al. Early clinical and angiographic outcomes after robotic-assisted coronary artery bypass surgery. J Thorac Cardiovasc Surg 2014;147:179-85. 26. Head SJ, Borgermann J, Osnabrugge RL, et al. Coronary artery bypass grafting: part 2—optimizing outcomes and future prospects. Eur Heart J 2013;34:2873-86.