Repeat Coronary Bypass Surgery or Percutaneous Coronary Intervention After Previous Surgical Revascularization

ORIGINAL ARTICLE

Repeat Coronary Bypass Surgery or Percutaneous Coronary Intervention After Previous Surgical Revascularization Chaim Locker, MD; Lawrence E. Greiten, MD; Malcolm R. Bell, MD; Robert L. Frye, MD; Amir Lerman, MD; Richard C. Daly, MD; Kevin L. Greason, MD; Sameh M. Said, MD; Brian D. Lahr, MS; John M. Stulak, MD; Joseph A. Dearani, MD; and Hartzell V. Schaff, MD Abstract Objective: To assess long-term survival with repeat coronary artery bypass grafting (RCABG) or percutaneous coronary intervention (PCI) in patients with previous CABG. Methods: From January 1, 2000, through December 31, 2013, 1612 Mayo Clinic patients underwent RCABG (n¼215) or PCI (n¼1397) after previous CABG. The RCABG cohort was grouped by use of saphenous vein grafts only (n¼75), or with additional arterial grafts (n¼140); the PCI cohort by, bare metal stents (BMS; n¼628), or drug-eluting stents (DES; n¼769), and by the treated target into native coronary artery (n¼943), bypass grafts only (n¼338), or both (n¼116). Multivariable regression and propensity score analysis (n¼280 matched patients) were used. Results: In multivariable analysis, the 30-day mortality was increased in RCABG versus PCI patients (hazard ratio [HR], 5.32; 95%CI, 2.34-12.08; P<.001), but overall survival after 30 days improved with RCABG (HR, 0.72; 95% CI, 0.55-0.94; P¼.01). Internal mammary arteries were used in 61% (129 of 215) of previous CABG patients and improved survival (HR, 0.82; 95% CI, 0.69-0.98; P¼.03). Patients treated with drug-eluting stent had better 10-year survival (HR, 0.74; 95% CI, 0.59-0.91; P¼.001) than those with bare metal stent alone. In matched patients, RCABG had improved late survival over PCI: 48% vs 33% (HR, 0.57; 95% CI, 0.35-0.91; P¼.02). Compared with RCABG, patients with PCI involving bypass grafts (n¼60) had increased late mortality (HR, 1.62; 95% CI, 1.10-2.37; P¼.01), whereas those having PCI of native coronary arteries (n¼80) did not (HR, 1.09; 95% CI, 0.75-1.59; P¼.65). Conclusion: RCABG is associated with improved long-term survival after previous CABG, especially compared with PCI involving bypass grafts. ª 2019 Mayo Foundation for Medical Education and Research

P

atients with previous coronary artery bypass grafting (CABG) who develop recurrent ischemic symptoms present a clinical challenge.1 Treatment has evolved over the past decades with improved medical management, the development of percutaneous coronary interventions (PCI), and evolution of CABG surgery.2 Patients who require repeat CABG (RCABG) have, in general, a higher risk for reoperation because of older age, comorbidities, and progression of native coronary artery and vein graft disease. In addition, there may be lack of suitable conduits and

n

Mayo Clin Proc. 2019;94(9):1743-1752

increased risk of injury to functioning bypass grafts during sternal reentry.3 In addition, symptomatic post-CABG patients often have coronary vessel disease limited to one or two territories. Because of the challenges of RCABG and the often limited disease, PCI is generally preferred for repeat revascularization.4 In a report from the National Cardiovascular Data Registry,5 patients with previous CABG represented 17.5% of all patients undergoing PCI. Further, the rate of RCABG has decreased dramatically. In one center,6

Mayo Clin Proc. n September 2019;94(9):1743-1752 n https://doi.org/10.1016/j.mayocp.2019.01.048 www.mayoclinicproceedings.org n ª 2019 Mayo Foundation for Medical Education and Research

From the Department of Cardiovascular Surgery (C.L., L.E.G., R.C.D., K.L.G., S.M.S, J.M.S., J.A.D., H.V.S.), Department of Cardiovascular Medicine (M.R.B., A.L., R.L.F.), and Division of Biomedical Statistics and Informatics (B.D.L.), Mayo Clinic, Rochester, MN.

1743

MAYO CLINIC PROCEEDINGS

RCABG accounted for 7.2% of patients having surgical revascularization from 1990 to 1994 but only 2.2% from 2005 to 2009. Given that operative mortality after RCABG is increased two- to four-fold compared with the primary operation, the indications are relatively strict.7 According to the 2014 European Society of Cardiology/European Association of Cardiothoracic Surgery guidelines on myocardial revascularization, RCABG should be considered for patients with diffusely diseased or occluded bypass grafts, multiple totally occluded native coronary arteries, absence of patent arterial grafts, or reduced systolic left ventricular function.8 Furthermore, the executive summary of the guidelines for CABG surgery from the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines states that it is reasonable to proceed with RCABG in symptomatic patients with left main disease, three-vessel disease, or a patent left internal mammary artery (IMA) to the left anterior descending coronary artery and ischemia in the distribution of the right or left circumflex coronary arteries.9 The purpose of the study was to review our outcomes with contemporary revascularization methods from January 1, 2000, through December 31, 2013, in all consecutive patients with multivessel disease who required repeat coronary revascularization after previous CABG, an era when stents and distal embolic protection were widely usable in PCI, compared with the modern surgical techniques. We hypothesized that RCABG may confer a long-term survival benefit compared with PCI. METHODS Study Design and Patients After approval by the Mayo Clinic Institutional Review Board, data were collected from our computerized cardiac surgery database, our percutaneous transluminal coronary angioplasty registry, and the patients’ electronic health records. Patient data and definitions followed the Society of Thoracic Surgeons National Cardiac Surgery Database and the American 1744

Mayo Clin Proc.

n

College of Cardiology database definitions and guidelines. Early mortality was defined in both cohorts as mortality occurring during the hospitalization in which the procedure was performed, as well as mortality occurring after hospital dismissal but within 30 days of the procedure. Late mortality was defined as any death that was not early mortality. Follow-up information was obtained via telephone calls and questionnaires mailed to patients at regular intervals and through review of the patients’ electronic records and the Social Security Death Index, supplemented by information provided from Accurint for Health Care (LexisNexis Risk Solutions, Alpharetta, Georgia) in both cohorts. From January 1, 2000, through December 31, 2013, a total of 1612 patients underwent revascularization with either RCABG (n¼215) or PCI (n¼1397) after having a previous CABG. The RCABG cohort was grouped into patients treated with arterial grafts (IMA and/or radial artery) with or without the addition of saphenous vein grafts (SVGs) (n¼140), or patients treated with SVG only (n¼75). The PCI cohort was grouped by the type of intervention performed, including bare metal stent (BMS; n¼628) or drug-eluting stent (DES; n¼769). PCI patients were further stratified by the type of the treated target into the native coronary artery (n¼943), bypass grafts only (n¼338), or both (n¼116). To assess each candidate risk factor for a nonproportional effect over early (0-30 days) and late (>30 days) follow-up, a time-dependent variable distinguishing the two time intervals was used to construct time-by-risk-factor interactions in Cox regression. Nonproportional effects were screened for importance on the basis of those interactions retained from stepwise variable selection with backwards elimination (a¼0.05), and thus express a relative hazard of death that differs between early and late follow-up. Indications for repeat myocardial revascularization were based on standard clinical and angiographic criteria.10 Surgical techniques for RCABG were previously described and followed.11-13

September 2019;94(9):1743-1752

n

https://doi.org/10.1016/j.mayocp.2019.01.048 www.mayoclinicproceedings.org

REPEAT CABG VS PCI

TABLE 1. Baseline Characteristics in Patients Treated With RCABG or PCI Before and After Propensity Score Matching Before matching RCABG (n¼215)a

Variable Age, y

69.7 (62.0-74.6)

Male

70.6 (62.5-77.6)

181 (84.2)

BMI (kg/m2)d

1095 (78.4)

29.0 (26.3-32.7) d

Diabetes mellitus d

Hypertension

29.3 (26.2-33.0)

65 (30.2)

547 (39.3)

177 (82.3)

d

Renal failure PVDd MI

PCI (n¼1397)a

d

1177 (86.0)

CHF

COPD

CVA/TIAd d

Left main disease EF  40

P

.02

b

70.2 (63.4-75.6)

69.0 (64.1-75.7)

.88b

.05

c

115 (82.1)

112 (80.0)

.65c

29.1 (25.9-32.7)

29.2 (26.7-32.6)

.53b

.44b .01

c

47 (33.6)

43 (30.7)

.61c

.16

c

114 (82.9)

110 (80.3)

.58c

c

5 (3.6)

3 (2.2)

.48c

28 (20.0)

24 (18.3)

.73c

70 (50.0)

65 (47.4)

.67c

64 (4.6)

.37

46 (21.4)

272 (19.9)

.61c

606 (44.6)

.25

c

269 (20.4)

<.001

18 (12.9)

22 (17.2)

.32c

23 (10.7)

170 (12.4)

.48

c

14 (10.0)

12 (9.0)

.78c

26 (13.9)

200 (14.6)

.79c

77 (5.5)

<.001

21 (9.8) d

PCI (n¼140)a

7 (3.3) 105 (48.8)

d

After matching RCABG (n¼140)a

P

97 (45.1) 45 (20.9)

181 (13.0)

c

19 (14.5)

19 (15.0)

.92c

c

41 (29.3)

32 (22.9)

.22c

c

24 (17.1)

27 (19.3)

.64c

.002

<.001

c

Diseased vessels 1 2 3 Elective procedured Year of procedure d

IMA, prior CABG

4 (1.9) 30 (14.0) 181 (84.2)

508 (36.4) 609 (43.6) 280 (20.0)

50 (23.3)

691 (49.5)

2003 (2001-07) 129 (60.8)

<.001c

2005 (2002-09)

<.001

1092 (84.1)

<.001

b c

.99c 4 (2.9) 28 (20.0) 108 (77.1)

4 (2.9) 29 (20.7) 107 (76.4)

40 (28.6)

44 (31.4)

2003 (2001-07)

2003 (2001-06)

89 (65.0)

89 (65.0)

.60c .51b >.99c

a

Values are count (percentage) for categorical variables and median (IQR) for continuous variables. Wilcoxon rank sum test for unmatched comparisons; Wilcoxon signed-rank test for matched pair comparisons. c Pearson c2 test for unmatched comparisons; McNemar test for matched pair comparisons. d Data were missing for a small proportion of patients (6% for prior use of IMA, 5% for CHF, 4% for all other variables) in the original sample; missing data tended to be proportionally higher (15% for CVA/TIA, 9% for CHF, 6% for all other variables) in the matched sample due to matched pair deletion even when missing in only one of the two patients. b

BMI ¼ body mass index; CABG ¼ coronary artery bypass grafting; CHF ¼ congestive heart failure; COPD ¼ chronic obstructive pulmonary disease; CVA ¼ cerebrovascular accident; EF ¼ ejection fraction; IMA ¼ internal mammary artery; IQR ¼ interquartile range; MI ¼ myocardial infarction; PCI ¼ percutaneous coronary intervention; PVD ¼ peripheral vascular disease; RCABG ¼ repeat coronary artery bypass grafting; TIA ¼ transient ischemic attack.

Statistical Analysis Descriptive statistics on baseline characteristics are presented as absolute numbers and percentages for discrete variables and as median and interquartile range for continuous variables. Baseline differences between treatment groups were determined by the Pearson c2 test or Wilcoxon rank sum test, as appropriate. Kaplan-Meier survival curves were estimated to show the proportion of patients who survived as a function of time. A modification of the standard Kaplan-Meier estimator for median survival, via reversal of the censor and death codes, was used to estimate median follow-up time. All analyses were performed using SAS statistical programming language (Version 9.4, SAS Institute Inc, Cary, NC). Mayo Clin Proc. n September 2019;94(9):1743-1752 www.mayoclinicproceedings.org

n

Cox proportional hazards (PH) regression was used to analyze the association of treatment group with survival time before and after adjustment for potential confounding variables. The validity of the PH model assumption was assessed by residual plots, as well as by statistical testing. In particular, a time-dependent variable distinguishing early (0-30 days) and late (>30 days) intervals of follow-up was constructed for the purpose of testing a time-bytreatment interaction. A significant interaction indicated evidence of a differential influence of treatment over time, which was represented by separate estimates of hazard for early and late death. Primary analyses were conducted on the overall groups using multivariable Cox PH regression to assess the independent

https://doi.org/10.1016/j.mayocp.2019.01.048

1745

MAYO CLINIC PROCEEDINGS

Survival, %

100

76%

80 52%

60

73%

40

20%

44%

20

Group by time interaction P<.001

0 0

10

15

60 (73) 309 (588)

9 (110) 60 (712)

5

No. at risk (deaths) Repeat CABG 215 (-) PCI 1,397 (-)

13%

Follow-up, y 114 (41) 809 (323)

A

Survival, %

100

78%

80 48%

60 70%

40

16%

20

33%

Group by time interaction P=.05

0 0

No. at risk (deaths) Repeat CABG 140 (-) PCI 140 (-)

13%

10

15

35 (51) 29 (74)

6 (71) 4 (87)

5 Follow-up, y 76 (25) 80 (37)

B

Survival, %

100

78% 71%

80

48%

60

20 0

23%

69%

40

13%

38%

Group effect P=.04 0

No. at risk (deaths) Repeat CABG 140 (-) PCI, native 80 (-) artery only PCI, any graft 60 (-)

27%

5

10

15

76 (25) 44 (21)

35 (51) 17 (38)

6 (71) 4 (44)

36 (16)

12 (36)

0 (43)

Follow-up, y

C FIGURE 1. (A) Kaplan-Meier curve for overall long-term survival of repeat coronary artery bypass grafting (RCABG) vs percutaneous coronary intervention (PCI) in unmatched patients. For 0-30 days: RCABG: PCI hazard ratio (HR), 4.51, 95% CI, 2.02-10.04; P<.001; for >30 days: RCABG: PCI HR, 0.92; 95% CI, 0.75-1.13; P¼0.411. Error bars indicate 95% CIs. (B) Kaplan-Meier curve for overall survival of propensity scoreematched patients treated with RCABG vs PCI. For 0-30 days: RCABG: PCI HR, 3.00, 95% CI, 0.61-14.86; P¼0.178; for >30 days: RCABG: PCI HR, 0.57, 95% CI, 0.35-0.91; P¼.02. Error bars indicate 95% CIs. (C) Kaplan-Meier curve for overall survival of propensity scoreematched patients treated with RCABG vs PCI. The latter group was divided into PCI of the native coronary arteries only vs PCI of bypass grafts. RCABG: PCI of native coronary arteries HR, 1.09, 95% CI, 0.75-1.59; P¼0.652; RCABG: PCI of bypass grafts HR, 1.62, 95% CI, 1.10-2.37; P¼.014.

1746

Mayo Clin Proc.

n

association of treatment, and other predictor variables, with survival time. In addition to treatment group (RCABG vs PCI), the model included: age, sex, body mass index, year of intervention, diabetes mellitus, hypertension, renal failure, peripheral vascular disease, myocardial infarction, congestive heart failure, pulmonary disease, cerebrovascular disease, ejection fraction of 40% or less, left main disease, number of diseased vessels, intervention status (nonelective vs elective), and use of IMA conduit in previous CABG. Missing data were imputed with the median value or modal category, as appropriate, and continuous predictors were modeled using restricted cubic splines to allow for nonlinear relationships. As an initial modeling step, predictor variables were collectively screened for time-bypredictor interactions (ie, non-PH) using step-down backwards variable selection and were retained in the final model if significant. For this and all other analyses, statistical significance was defined by P value less than or equal to .05. As an additional method for treatment comparison, we used propensity analysis to model and compensate for confounding effects of nonrandom treatment selection. The same covariates in the multivariable analysis were used to fit the propensity model, with use of flexible regression methods to allow for nonlinear relationships (ie, splines) and missing data. The latter was achieved by setting the missing predictor values to constants and adding missingness indicator variables to the model, thereby incorporating missingness patterns into the adjustment. The propensity model used logistic regression to convert all the input variables into a single score measuring the predicted log odds of receiving RCABG (instead of PCI). Caliper matching based on the optimal algorithm was then used to sample equally sized subgroups from the two treatments, matching patients treated with PCI to patients who had RCABG on propensity score within a tolerance of 0.2 of an SD. The adequacy of the propensity score in minimizing the effect of treatment selection bias was shown by the comparison of individual covariates between matched

September 2019;94(9):1743-1752

n

https://doi.org/10.1016/j.mayocp.2019.01.048 www.mayoclinicproceedings.org

REPEAT CABG VS PCI

groups and standardized differences before and after matching. A stratified Cox model was then used in the final step to evaluate the association of group with survival time among RCABG- and PCI-matched patients.

TABLE 2. Multivariable Analysis of Risk Factors for Survival (All Patients) Risk factor

HR (95% CI)

P value

Age (IQR effect: 77 vs 62 years)

2.32 (1.94-2.77)

<.001

Sex, male vs female

0.99 (0.84-1.18)

.94

2

RESULTS Baseline characteristics of patients treated with RCABG or PCI are shown in Table 1. Compared with patients undergoing PCI, RCABG patients were more often men and more frequently had clinically important stenosis of the left main coronary artery and three-vessel coronary artery disease, and ejection fraction of 40% or less. Patients undergoing PCI had an increased frequency of diabetes mellitus, congestive heart failure, more frequent IMA grafts from a previous CABG, and were less likely to have undergone an elective procedure. Unadjusted Outcomes During a median follow-up period of 10.4 years (25th to 75th percentile: 5.7-14.4 years), there were 837 deaths among the 1612 patients. In unadjusted analyses, RCABG patients appeared to have worse survival over early follow-up (30-day mortality: 4.7% vs 1.1 %) yet a more favorable longterm prognosis than PCI patients (median survival: 10.3 vs 9.0 years). The crossing of the survival curves during follow-up (Figure 1A) shows a possible interaction between treatment and time, and the test for interaction (P<.001) confirmed a difference in hazard of death between early and late follow-up. In unadjusted analysis, RCABG was associated with an increased risk of dying early hazard ratio (HR), 4.51; 95% CI, 2.02-10.04; P<.001) and no survival benefit thereafter (HR, 0.92; 95% CI, 0.751.13; P¼.41). Adjusted Outcomes Multivariable Model. A similar time-varying treatment effect was shown after multivariable adjustment for potential confounding factors (interaction, P<.001); RCABG was associated with a higher risk of early death (adjusted HR, 5.32; 95% CI, 2.34-12.08; P<.001) but a 28% reduction in risk of late Mayo Clin Proc. n September 2019;94(9):1743-1752 www.mayoclinicproceedings.org

n

BMI (IQR effect: 33 vs 26 kg/m )

1.08 (0.96-1.21)

.03

Diabetes mellitus, yes vs noa

1.65 (1.42-1.91)

<.001

Hypertension, yes vs noa

1.24 (1.00-1.53)

.05

Renal failure, yes vs noa

2.29 (1.73-3.05)

<.001

PVD, yes vs noa

1.54 (1.30-1.82)

<.001

a

1.18 (1.02-1.36)

.03

CHF, yes vs noa

1.67 (1.40-2.00)

<.001

COPD, yes vs noa

1.49 (1.23-1.81)

<.001

CVA/TIA, yes vs noa

1.20 (1.00-1.45)

.05

Left main disease 50% vs <50%

1.12 (0.88-1.44)

.35

EF 40% vs >40%

1.72 (1.42-2.07)

<.001

Diseased vessels 1 2 3

1.0 (referent) 1.07 (0.90-1.27) 1.34 (1.09-1.64)

Status, urgent/emergent vs electiveb

0.88 (0.76-1.01)

.07

Calendar year (IQR effect: 2009 vs 2002)b

1.11 (0.89-1.38)

.47

IMA used in previous CABG, yes vs noa

0.82 (0.69-0.98)

.03

5.32 (2.34-12.08) 0.72 (0.55-0.94)

Interaction, P<.001 <.001 .02

MI, yes vs no

Intervention: RCABG vs PCIb Early effect: 0 e 30 days Late effect: >30 days

.01

a

Single imputation was used to avoid case wise deletion due to missing data (we set continuous variables to the median value and discrete variables to the most frequent category). b To assess each risk factor for a nonproportional effect over early (0-30 days) and late (>30 days) follow-up, a time-dependent variable was used to construct time-by-risk factor interactions in Cox regression. Noneproportional hazards (PH) effects were screened for importance based on those interactions retained from stepwise variable selection with backwards elimination (a level of 0.05). Non-PH treatment effect in this model thus expresses a relative hazard of death that changes over follow-up. BMI ¼ body mass index; CABG ¼ coronary artery bypass grafting; CHF ¼ congestive heart failure; COPD ¼ chronic obstructive pulmonary disease; CVA ¼ cerebrovascular accident; EF ¼ ejection fraction; HR ¼ hazard ratio; IMA ¼ internal mammary artery; MI ¼ myocardial infarction; PCI ¼ percutaneous coronary intervention; PVD ¼ peripheral vascular disease; TIA ¼ transient ischemic attack.

death (adjusted HR, 0.72; 95% CI, 0.55-0.94; P¼.01) compared with PCI. Additional risk factors for increased mortality identified in this multivariable analysis are presented in Table 2. Importantly, the use of an IMA conduit in a previous CABG was independently associated with reduced overall mortality (adjusted HR, 0.82; 95% CI, 0.69-0.98; P¼.03). When the multivariable Cox analysis was repeated on the subset of 215 RCABG patients, of whom 140 (65.1%) had arterial grafts at reoperation (of these, 66 [47.1] had

https://doi.org/10.1016/j.mayocp.2019.01.048

1747

MAYO CLINIC PROCEEDINGS

IMA grafts at their previous operation) and 75 (35%) had SVG only (of these, 63 [84%] had IMA grafts at their previous operation), there was no overall difference in adjusted survival according to type of grafts at reoperation (Table 3). However, there was a trend toward a significant interaction between grafts involved at previous and repeat CABG (P¼.07), which suggested a modest (nonsignificant) survival improvement with arterial grafts versus SVG alone in patients who had received an IMA graft in their previous CABG (adjusted HR, 0.64, 95% CI 0.371.10; P¼.10). A separate multivariable Cox analysis was also performed on the subset of patients treated with PCI to investigate differences in adjusted survival by treatment modality. Compared with patients treated with BMS, risk of death was significantly reduced in patients receiving DES (adjusted HR, 0.74; 95% CI, 0.59-0.91; P¼.005) (Table 3). Propensity Score Matching Covariate balance after matching of 140 pairs treated with RCABG and PCI resulted in even distribution of baseline characteristics (Table 1). Graphic depictions presented in Figure 2 further show the statistical

comparability between matched groups, with minimal differences noted between the two overlaid propensity distributions and the standardized effects of all baseline variables. Reduction in treatment selection bias as a result of propensity score matching ranged from 40% to 132% among the covariates that initially showed baseline differences. As shown in Figure 1B, a differential pattern of early versus late survival was borderline significant in the matched groups (interaction, P¼.05). Although not statistically significant, patients treated with RCABG had a three-fold higher mortality risk in the first 30 days compared with matched patients treated with PCI (HR, 3.00; 95% CI, 0.6114.86; P¼.18). In contrast, RCABG was associated with a significant survival benefit after 30 days, corresponding to a 43% lower hazard of death than PCI (HR, 0.57; 95% CI, 0.350.91; P¼.02). The two survival curves crossed at about 1 year of follow-up, after which RCABG survival rates were higher compared with PCI for most of the follow-up period (10-year survival: 48% vs 33%). However, later in the period the differences between the two curves began to narrow and the curves had crossed back over.

TABLE 3. Association of PCI and RCABG Treatment Subtypes With Long-term Mortality No. (%) Total RCABG group (n¼215) Graft type, main effect Arterial (SVG) SVG only Graft type, interaction effect No IMA in previous CABG (n¼83b) Arterial (SVG) SVG only

140 (65.1) 75 (34.9)

71 (85.5) 12 (14.5)

Unadjusted HR (95% CI)

Adjusteda HR (95% CI)

P¼.61 0.90 (0.61-1.33) 1.0 (referent) Interaction, P¼.04

P¼.48 0.85 (0.55-1.33) 1.0 (referent) Interaction, P¼.07

1.63 (0.73-3.65) 1.0 (referent)

1.60 (0.68-3.76) 1.0 (referent)

IMA in previous CABG (n¼129b) Arterial (SVG)

66 (51.2)

0.61 (0.37-1.01)

0.64 (0.37-1.10)

SVG only

63 (48.8)

1.0 (referent)

1.0 (referent)

P<.001 0.72 (0.61-0.85) 1.0 (referent)

P¼.005 0.74 (0.59-0.91) 1.0 (referent)

Total PCI group (n¼1397) Stent type DES (with/without BMS) BMS only a

769 (55.0) 628 (44.9)

Model adjusted for same baseline covariates as listed in Table 2. Numbers for these subgroups do not add up to 215 because three patients whose IMA status at previous CABG was unknown.

b

BMS ¼ bare-metal stent; CABG ¼ coronary artery bypass grafting; DES ¼ drug-eluting stent; HR ¼ hazard ratio; IMA ¼ internal mammary artery; PCI ¼ percutaneous coronary intervention; RCABG ¼ repeat coronary artery bypass grafting; SVG ¼ saphenous vein graft.

1748

Mayo Clin Proc.

n

September 2019;94(9):1743-1752

n

https://doi.org/10.1016/j.mayocp.2019.01.048 www.mayoclinicproceedings.org

REPEAT CABG VS PCI

20

20 PCI

Repeat CABG

PCI

15

15

N=280 Percent

Percent

N=1,612 10

10

5

5

0

0 –6

A

Repeat CABG

–4 –2 0 2 Propensity scores, original cohort

–6

4

B

Propensity Score (logit) No. diseased vessels 1 2 3 Left main disease ≥50% Non-elective status IMA used in prior CABG Year of intervention CHF EF ≤40% Diabetes Gender Age Hypertension MI Renal failure COPD BMI PVD CVA/TIA

4

Before PS-matching After PS-matching –1.0

C

–4 –2 0 2 Propensity scores, matched sample

0.0 1.0 Standardized difference

1.5

FIGURE 2. Baseline comparability of repeat coronary artery bypass grafting (RCABG) vs percutaneous coronary intervention (PCI) groups before and after propensity score matching. (A) Histogram of propensity score distribution in original (unmatched) cohort. (B) Histogram of propensity scores among matched pairs. (C) Standardized differences before and after matching. BMI ¼ basal metabolic index; CHF ¼ congestive heart failure; COPD ¼ chronic obstructive pulmonary disease; CVA/TIA ¼ cerebrovascular accident/ transient ischemic accident; EF ¼ ejection fraction; IMA ¼ internal mammary artery; MI ¼ myocardial infarction; PS ¼ propensity score; PVD ¼ peripheral vascular disease.

For groups treated with RCABG, PCI involving bypass grafts, and PCI involving only the native coronary arteries, there was no evidence of a differential pattern between early and late follow-up (interaction, P¼.22); therefore, relative survival differences were averaged over the entire followup period. This analysis revealed an overall Mayo Clin Proc. n September 2019;94(9):1743-1752 www.mayoclinicproceedings.org

n

survival difference among the three groups (P¼.04). In particular, compared with RCABG, PCI involving bypass grafts had increased risk of mortality (HR, 1.62; 95% CI, 1.10-2.37; P¼.01), whereas PCI targeting only the native coronary arteries did not (HR, 1.09; 95% CI, 0.75-1.59; P¼.65) (Figure 1C).

https://doi.org/10.1016/j.mayocp.2019.01.048

1749

MAYO CLINIC PROCEEDINGS

DISCUSSION Most previous investigations comparing outcome of patients undergoing repeat revascularization with PCI or RCABG suggest increased early mortality among RCABG patients but similar long-term survival.14,15 In this study, which includes a contemporary cohort, the risk for early mortality was also increased for RCABG versus PCI. However, long-term survival of patients undergoing RCABG was significantly better than that of patients treated with PCI in both the multivariable analysis and in propensity-matched patients. Compared with patients treated with PCI, the hazard of late death was reduced by 28% in patients having RCABG; survival at 10 years after RCABG reflected a 15% absolute improvement over PCI. These results contrast those of Weintraub et al16 who reported that initial mortality was higher after RCABG, but after adjusting for differences in baseline characteristics, they found no significant difference in long-term outcomes. However, in the most recent study by Brener et al,17 overall survival of unmatched and propensity matched patients favored RCABG. Our study supports this finding within a more contemporary cohort of patients. A possible explanation for reduced survival of patients with previous CABG having PCI may be related to the target vessel treated. In our study, patients who had exclusively PCI of the native coronary arteries had survival rates similar to those of patients undergoing RCABG. In contrast, patients who had PCI of SVGs, with or without treatment of native coronary arteries, had significantly worse survival rates than patients having RCABG. This finding is consistent with a recent report by Brilakis et al.18 In a study of 11,118 patients that compared PCI of bypass grafts to treatment of native coronary arteries, they found that PCI of bypass grafts was associated with significantly higher mortality (HR, 1.30), myocardial infarction (HR, 1.61) and need for repeat revascularization (HR, 1.60). Furthermore, in a recent randomized controlled study, Brilakis et al19 showed that 1750

Mayo Clin Proc.

n

in patients undergoing stenting of SVG lesions, no differences in outcomes between BMS or DES during 1 year of follow-up were found. Another recent study by Colleran et al20 found that in treatment of SVGs the superiority of DES over BMS shown at 1 year was lost at 5-year follow-up due to increased attrition of efficacy in the DES group. In our study, overall survival was best in those treated with a DES among patients treated with PCI, and this result was independent of other risk factors, including target vessel treated. These findings are consistent with the results of a systematic meta-analysis by Testa et al21 who reported similar risks of major adverse events and myocardial infarction in patients with SVG disease treated with a BMS or DES, but a trend toward a higher risk of death (odds ratio, 1.32) in patients receiving a BMS. Importantly, in contrast with previous reports in which a DES in a SVG was associated with increased risk of long-term death and stent thrombosis,22 our results showed that using a DES was relatively safe.23 Additionally, it appears that independent of the treatment type and any other risk factors, the presence of an IMA from the previous CABG was associated with significantly improved survival. Some clinicians believe that PCI is the preferred treatment for patients with patent left IMA grafts to the left anterior descending artery because of lack of additional survival benefit with RCABG.24 Our results support this strategy only if PCI is limited to native coronary arteries. In current practice, PCI is considered by many to be the first option for patients who have medically refractory symptoms after previous CABG. Among the subgroup of patients in the Angina With Extremely Serious Operative Mortality Evaluation (AWESOME) trial with previous CABG, both physicians and patients chose PCI rather than RCABG by a 2:1 margin.25 This practice was based largely on data that RCABG is associated with higher procedural risk than PCI. In that report, there was a nonsignificant tendency toward better survival of PCI patients than of patients having RCABG. However, our data from both

September 2019;94(9):1743-1752

n

https://doi.org/10.1016/j.mayocp.2019.01.048 www.mayoclinicproceedings.org

REPEAT CABG VS PCI

the overall cohort of patients and the propensity-matched subgroup show a clear overall survival benefit of RCABG. This potential survival advantage should be considered in the selection of revascularization treatment options after previous CABG, especially in patients who require intervention on diseased SVG. Study Limitations The study has several limitations. It is retrospective, observational, and from a single center. Although we compensated for known baseline differences among the study groups with propensity-scoreematched analysis and adjusted for known risk factors of overall survival with multivariable analysis, the conclusions may be limited by inherent selection bias and unmeasured confounding variables including frailty. Another limitation is lack of data on completeness of revascularization 2628 and on secondary prevention.29 We did not analyze subgroups of DES, but the date of PCI procedure had no influence on results, and it seems unlikely that stratification by type of DES would change the overall findings. CONCLUSION Overall long-term survival after previous CABG may be improved with RCABG compared with PCI, largely because of decreased survival of patients who undergo PCI of diseased grafts. The use of an IMA in the previous CABG is associated with improved survival. The data suggest that RCABG may be underused in current practice.30 Further studies are required to assess long-term outcomes of revascularization strategies. Finally, it seems important for guidelines to consider the need for intervention on SVG in view of improved outcomes with RCABG. Abbreviations and Acronyms: BMS = bare metal stent; CABG = coronary bypass grafting; DES = drug eluting stent; HR = hazard ratio; IMA = internal mammary artery; PCI = percutaneous coronary intervention; RCABG = repeat coronary artery bypass grafting; SVG = saphenous vein graft

Potential Competing Interests: The authors report no competing interests. Mayo Clin Proc. n September 2019;94(9):1743-1752 www.mayoclinicproceedings.org

n

Correspondence: Address to Chaim Locker, MD, Department of Cardiovascular Surgery, Mayo Clinic, 200 First St SW, Rochester, MN, 55905 ([email protected]).

REFERENCES 1. Weintraub WS, Jones EL, Craver JM, Grosswald R, Guyton RA. In-hospital and long-term outcome after reoperative coronary artery bypass graft surgery. Circulation. 1995;92(suppl 9):50-57. 2. Sabik JF 3rd, Blackstone EH, Houghtaling PL, Walts PA, Lytle BW. Is reoperation still a risk factor in coronary artery bypass surgery? Ann Thorac Surg. 2005;80(5):1719-1727. 3. Tector A. Invited commentary. Ann Thorac Surg. 2009;87(5): 1391. 4. Fosbol EL, Zhao Y, Shahian DM, Grover FL, Edwards FH, Peterson ED. Repeat coronary revascularization after coronary artery bypass surgery in older adults: The Society of Thoracic Surgeons’ national experience 1991e2007. Circulation. 2013; 127(16):1656-1663. 5. Brilakis ES, Rao SV, Banerjee S, et al. Percutaneous coronary intervention in native arteries versus bypass grafts in prior coronary artery bypass grafting patients: a report from the National Cardiovascular Data Registry. J Am Coll Cardiol Int. 2011;4(8): 844-850. 6. Spiliotopoulos K, Maganti M, Brister S, Rao V. Changing pattern of reoperative coronary artery bypass grafting: a 20-year study. Ann Thorac Surg. 2011;92(1):40-46. 7. Yap CH, Sposato L, Akowuah E, et al. Contemporary results show repeat coronary artery bypass grafting remains a risk factor for operative mortality. Ann Thorac Surg. 2009;87(5):13861391. 8. Authors/Task Force members, Windecker S, Kolh P, et al. 2014 ESC/EACTS guidelines on myocardial revascularization: The Task Force on Myocardial Revascularization of the European Society of Cardiology (ESC) and the European Association for Cardio-Thoracic Surgery (EACTS) developed with the special contribution of the European Association of Percutaneous Cardiovascular Interventions (EAPCI). Eur Heart J. 2014;35(37): 2541-2619. 9. Hillis LD, Smith PK, Anderson JL, et al. 2011 ACCF/AHA guideline for coronary artery bypass graft surgery, executive summary: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines. Circulation. 2011;124(23):2610-2642. 10. Akins CW, Buckley MJ, Daggett WM, et al. Reoperative coronary grafting: changing patient profiles, operative indications, techniques, and results. Ann Thorac Surg. 1994;58(2):359-364; discussion 364-365. 11. Loop FD, Thurer RL, Lytle BW, Cosgrove DM. Reoperation for myocardial revascularization. World J Surg. 1978;2(6):719-727. 12. Schaff HV, Orszulak TA, Gersh BJ, et al. The morbidity and mortality of reoperation for coronary artery disease and analysis of late results with use of actuarial estimate of event-free interval. J Thorac Cardiovasc Surg. 1983;85(4): 508-515. 13. Loop FD, Lytle BW, Cosgrove DM, et al. Reoperation for coronary atherosclerosis: changing practice in 2509 consecutive patients. Ann Surg. 1990;212(3):378-386. 14. Cole JH, Jones EL, Craver JM, et al. Outcomes of repeat revascularization in diabetic patients with prior coronary surgery. J Am Coll Cardiol. 2002;40(11):1968-1975. 15. Stephan WJ, O’Keefe JH Jr, Piehler JM, et al. Coronary angioplasty versus repeat coronary artery bypass grafting for patients with previous bypass surgery. J Am Coll Cardiol. 1996;28(5): 1140-1146. 16. Weintraub WS, Jones EL, Morris DC, King SB 3rd, Guyton RA, Craver JM. Outcome of reoperative coronary bypass surgery versus coronary angioplasty after previous bypass surgery. Circulation. 1997;95(4):868-877.

https://doi.org/10.1016/j.mayocp.2019.01.048

1751

MAYO CLINIC PROCEEDINGS

17. Brener SJ, Lytle BW, Casserly IP, Ellis SG, Topol EJ, Lauer MS. Predictors of revascularization method and long-term outcome of percutaneous coronary intervention or repeat coronary bypass surgery in patients with multivessel coronary disease and previous coronary bypass surgery. Eur Heart J. 2006; 27(4):413-418. 18. Brilakis ES, O’Donnell CI, Penny W, et al. Percutaneous coronary intervention in native coronary arteries versus bypass grafts in patients with prior coronary artery bypass graft surgery: insights from the Veterans Affairs Clinical Assessment, Reporting, and Tracking Program. J Am Coll Cardiol Int. 2016;9(9):884-893. 19. Brilakis ES, Edson R, Bhatt DL, et al. Drug-eluting stents versus bare-metal stents in saphenous vein grafts: a double blind, randomized trial. Lancet. 2018;391(10134):1997-2007. 20. Colleran R, Kufner S, Mehilli J, et al. Efficacy over time with drug-eluting stents in saphenous vein graft lesions. J Am Coll Cardiol. 2018;71(18):1973-1982. 21. Testa L, Agostoni P, Vermeersch P, et al. Drug eluting stents versus bare metal stents in the treatment of saphenous vein graft disease: a systematic review and meta-analysis. EuroIntervention. 2010;6(4):527-536. 22. Vermeersch P, Agostoni P, Verheye S, et al. Increased late mortality after sirolimus-eluting stents versus bare-metal stents in diseased saphenous vein grafts: results from the randomized DELAYED RRISC Trial. J Am Coll Cardiol. 2007;50(3):261-267. 23. Aggarwal V, Stanislawski MA, Maddox TM, et al. Safety and effectiveness of drug-eluting versus bare-metal stents in saphenous vein bypass graft percutaneous coronary interventions:

1752

Mayo Clin Proc.

n

24.

25.

26.

27.

28.

29.

30.

insights from the Veterans Affairs CART Program. J Am Coll Cardiol. 2014;64(17):1825-1836. Subramanian S, Sabik JF 3rd, Houghtaling PL, Nowicki ER, Blackstone EH, Lytle BW. Decision-making for patients with patent left internal thoracic artery grafts to left anterior descending. Ann Thorac Surg. 2009;87(5):1392-1398; discussion 1400. Morrison DA, Sethi G, Sacks J, et al. Percutaneous coronary intervention versus repeat bypass surgery for patients with medically refractory myocardial ischemia: AWESOME randomized trial and registry experience with post-CABG patients. J Am Coll Cardiol. 2002;40(11):1951-1954. Di Mauro M, Iaco AL, Contini M, et al. Reoperative coronary artery bypass grafting: analysis of early and late outcomes. Ann Thorac Surg. 2005;79(1):81-87. Xu B, Yang YJ, Han YL, et al. Validation of residual SYNTAX score with second-generation drug-eluting stents: one-year results from the prospective multicentre SEEDS study. Eurointervention. 2014;10(1):65-73. Azzalini L, Candilio L, Ojeda S, et al. Impact of incomplete revascularization on long-term outcomes following chronic total occlusion percutaneous coronary intervention. Am J Cardiol. 2018;121(10):1138-1148. Antoniades C, Bakogiannis C, Tousoulis D, et al. Preoperative atorvastatin treatment in CABG patients rapidly improves vein graft redox state by inhibition of Rac1 and NADPHoxidase activity. Circulation. 2010;122(suppl 11):S66-S73. Arsalan M, Mack MJ. Coronary artery bypass grafting is currently underutilized. Circulation. 2016;133(10):1036-1045.

September 2019;94(9):1743-1752

n

https://doi.org/10.1016/j.mayocp.2019.01.048 www.mayoclinicproceedings.org