Author + information
- Received October 21, 2016
- Revision received September 29, 2017
- Accepted October 10, 2017
- Published online December 11, 2017.
- Krishnan Ramanathan, MB, ChBa,∗ (, )
- James G. Abel, MDa,
- Julie E. Park, Mmathb,
- Anthony Fung, MBBSa,
- Verghese Mathew, MDc,
- Carolyn M. Taylor, MDa,
- G.B. John Mancini, MDa,
- Min Gao, MD, PhDb,
- Lillian Ding, MScd,
- Subodh Verma, MD, PhDe,
- Karin H. Humphries, DSca,b and
- Michael E. Farkouh, MD, MScf
- aUniversity of British Columbia, Vancouver, Canada
- bBC Centre for Improved Cardiovascular Health, Vancouver, Canada
- cLoyola University Medical Center and Stritch School of Medicine, Maywood, Illinois
- dCardiac Services British Columbia, Vancouver, Canada
- eSt. Michael’s Hospital, Toronto, Canada
- fPeter Munk Cardiac Centre and the Heart and Stroke Richard Lewar Centre, University of Toronto, Toronto, Canada
- ↵∗Address for correspondence:
Dr. Krishnan Ramanathan, University of British Columbia, 1081 Burrard St – B475, Vancouver, BC V6Z 1Y6, Canada.
Background Randomized trial data support the superiority of coronary artery bypass grafting (CABG) surgery over percutaneous coronary intervention (PCI) in diabetic patients with multivessel coronary artery disease (MV-CAD). However, whether this benefit is seen in a real-world population among subjects with stable ischemic heart disease (SIHD) and acute coronary syndromes (ACS) is unknown.
Objectives The main objective of this study was to assess the generalizability of the FREEDOM (Future REvascularization Evaluation in Patients with Diabetes Mellitus: Optimal Management of Multi-vessel Disease) trial in real-world practice among patients with diabetes mellitus and MV-CAD in residents of British Columbia, Canada. Additionally, the study evaluated the impact of mode of revascularization (CABG vs. PCI with drug-eluting stents) in diabetic patients with ACS and MV-CAD.
Methods In a large population-based database from British Columbia, this study evaluated major cardiovascular outcomes in all diabetic patients who underwent coronary revascularization between 2007 and 2014 (n = 4,661, 2,947 patients with ACS). The primary endpoint (major adverse cardiac or cerebrovascular events [MACCE]) was a composite of all-cause death, nonfatal myocardial infarction, and nonfatal stroke. The risk of MACCE with CABG or PCI was compared using multivariable adjustment and a propensity score model.
Results At 30-days post-revascularization, for ACS patients the odds ratio for MACCE favored CABG 0.49 (95% confidence interval [CI]: 0.34 to 0.71), whereas among SIHD patients MACCE was not affected by revascularization strategy (odds ratio: 1.46; 95% CI: 0.71 to 3.01; pinteraction <0.01). With a median follow-up of 3.3 years, the late (31-day to 5-year) benefit of CABG over PCI no longer varied by acuity of presentation, with a hazard ratio for MACCE in ACS patients of 0.67 (95% CI: 0.55 to 0.81) and the hazard ratio for SIHD patients of 0.55 (95% CI: 0.40 to 0.74; pinteraction = 0.28).
Conclusions In diabetic patients with MV-CAD, CABG was associated with a lower rate of long-term MACCE relative to PCI for both ACS and SIHD. A well-powered randomized trial of CABG versus PCI in the ACS population is warranted because these patients have been largely excluded from prior trials.
Globally, the prevalence of diabetes mellitus (DM) among adults is projected to reach 642 million by 2040 (1). Total deaths from DM are projected to rise by more than 50% in the next 10 years, thereby making DM and its complications important societal and public health concerns. Patients with DM are prone to a diffuse and rapidly progressive forms of atherosclerosis, which increases their likelihood of having multivessel coronary artery disease (MV-CAD) requiring revascularization (2). Selecting the optimal coronary revascularization strategy for patients with DM and MV-CAD is crucial to reduce the high rate of thrombotic complications and improve quality of life.
The choice between CABG and PCI in DM remains an area of intense discussion and debate. In 2012, the results of the randomized controlled FREEDOM (Future REvascularization Evaluation in Patients With Diabetes Mellitus: Optimal Management of Multi-vessel Disease) trial demonstrated lower rates of major adverse cardiovascular events in patients with stable ischemic coronary disease (SIHD) who were assigned to coronary artery bypass grafting (CABG) compared with percutaneous coronary intervention (PCI) using drug-eluting stents (PCI-DES), over a median of 3.8 years of follow-up (3). The results of FREEDOM also showed a borderline reduction in all-cause mortality (p = 0.049) favoring CABG over PCI-DES; this effect was confirmed in a robust meta-analysis of randomized trials (4).
Although carefully planned and conducted randomized trials provide the greatest intrinsic validity of the study question under consideration, the generalizability of these observations in the real world are often limited by the recruitment criteria. Therefore, well-conducted population-based analyses that provide complementary extrinsic validation of randomized trials remain important components of overall knowledge translation efforts and are often the sources of supplementary information for key stakeholders.
The main objective of our study was to assess the generalizability of the FREEDOM trial in real-world practice among patients with DM and MV-CAD in residents of British Columbia, Canada. Additionally, we evaluated the impact of mode of revascularization (CABG vs. PCI-DES) in diabetic patients with acute coronary syndromes (ACSs) and MV-CAD. Previous studies of CABG versus PCI in DM patients with MV-CAD consisted mostly of patients with stable ischemic heart disease (SIHD) (4). The present study allowed us to analyze the ACS subgroup that represents a large proportion of people with DM who are undergoing revascularization, and where there is guideline equipoise (5).
Cardiac Services British Columbia holds a province-wide registry that captures patients’ demographic and clinical data for all coronary angiography and revascularization procedures in British Columbia. This registry was used to define the study cohort. Hospitalization separation data for all patients hospitalized in British Columbia, including admission and discharge date, hospital identification code, and International Classification of Disease-10th Revision diagnosis codes, were used to identify the following outcomes: myocardial infarction (MI) and stroke. The Vital Statistics Deaths registry provides all death dates in British Columbia and was used to identify all-cause mortality. There are 5 cardiac catheterization sites in British Columbia, all with PCI and CABG capabilities. All 5 sites are located in the southern and central regions of British Columbia and are reliant on a “hub and spoke” referral process and an extensive transport network. In British Columbia, approximately 80% of all PCIs are carried out as ad hoc procedures with high rates of DES use (80%). We obtained ethics approval from the University of British Columbia-Providence Health Care Research Ethics Board.
This is a population-based, retrospective cohort study of all patients older than 20 years of age with DM and angiographically confirmed MV-CAD (stenosis of >70% in 2 or more major epicardial vessels, excluding the left main coronary artery) who underwent either PCI or isolated CABG between October 1, 2007 and January 31, 2014 in British Columbia. As in the FREEDOM trial, subjects with severe heart failure (New York Heart Association functional class III or IV), prior CABG or PCI within 6 months, prior valve surgery, 2 or more chronic total occlusions, ST-segment elevation MI within 72 h, and stroke within 6 months were excluded (Figure 1).
The primary outcome was the first occurrence of a major adverse cardiac or cerebrovascular event (MACCE), defined as a composite of all-cause mortality, nonfatal MI (International Classification of Disease-Tenth Revision [ICD-10] codes I21, I22) and nonfatal stroke (ICD-10 codes I60 to I64, H356, H341, H342, and H348) after revascularization. Secondary outcomes included the individual components of MACCE, repeat revascularization post-discharge (RR), and a composite of MACCE and repeat revascularization (MACCE[r]). Staged PCI was defined as a nonemergency PCI that was planned and performed within 2 months of the previous PCI; these PCIs were not included in identifying repeat revascularizations post-discharge. The validity of ICD-10 codes for determination of outcomes has been well validated (6).
Baseline variables were summarized as frequency and percentage for categorical variables and as mean and SD for continuous variables. The comparisons by revascularization strategy (PCI vs. CABG) and clinical presentation (ACS vs. SIHD) were tested using the chi-square test for categorical variables and the Student’s t-test for continuous variables. The proportional hazard assumption was not met over the 5-year follow-up period; therefore the analysis was divided into early (30-day) and late (31-day to 5-year) follow-up post-revascularization, with results reported separately for each time frame. For the first 30-days, event rates, odds ratios (ORs), and 95% confidence intervals (CIs) are reported for CABG versus PCI. For the late period, event rates are reported as Kaplan-Meier estimates, and Cox proportional hazard models were fitted to obtain hazard ratios (HRs) and 95% CIs for CABG versus PCI. Data were censored if a patient did not experience an outcome, had another revascularization procedure within 1 year, or reached the end of follow-up, which was March 31, 2014. When the individual outcomes were studied, death was additionally censored. To assess the impact of revascularization strategy on patients presenting with ACS compared with SIHD, the interaction term of acuity subtype by revascularization strategy was included in the model. To account for baseline differences, fully adjusted logistic and Cox models of MACCE were also constructed. Baseline characteristics with a p value < 0.20 in the univariate models were considered as potential confounders; factors with a p value < 0.05 remained in the final model, except for sex, which remained in the model regardless of its significance. Given the difference in baseline variables between PCI and CABG groups, a propensity score (PS) method, inverse probability of treatment weight (IPTW) (7), was performed as sensitivity analysis. The list of variables that were used to create the PS were age, fiscal year of procedure, hospital, presentation acuity (ACS vs. SIHD), DM type, left ventricular ejection fraction (LVEF), hypertension, history of congestive heart disease, pulmonary disease, peripheral arterial disease, renal insufficiency, liver or gastrointestinal disease, and extent of CAD. The use of stabilized weights provides more precise estimation of treatment effect without losing any observations from incomplete matching. Standardized differences were calculated to assess the balance in baseline variables between groups before and after applying IPTW. In addition, a multilevel hospital-by-procedure year variable was chosen as an instrumental variable (IV); IV analysis on 30-day MACCE was performed to obtain an OR estimate for CABG versus PCI while controlling for selection bias from unmeasured confounders. All analyses were performed using SAS software version 9.4 (SAS Institute, Cary, North Carolina).
From October 1, 2007 through January 31, 2014, 51,203 coronary revascularization procedures were performed in British Columbia: 40,053 PCIs (78%) and 11,150 isolated CABG procedures. Of this total, 4,819 coronary revascularization procedures (PCI or isolated CABG) were performed in 4,661 patients with DM and CAD who met the inclusion or exclusion criteria outlined in Figure 1.
The baseline characteristics of the overall cohort are described by revascularization strategy in Table 1. Patients who underwent CABG were significantly younger, less likely to be female, and less likely to present with ACS, and they presented with less peripheral arterial disease, pulmonary disease, and left ventricular dysfunction. Conversely, CABG-treated patients had more triple-vessel disease and more proximal left anterior descending coronary artery involvement.
When baseline characteristics are described by acuity of presentation—ACS versus SIHD—patients presenting with ACS (63%) were slightly older, more likely to be female and had more renal insufficiency and left ventricular dysfunction. The 2 groups did not differ with respect to triple-vessel disease or left anterior descending coronary artery involvement (Table 2), respectively. Among ACS patients, 65.2% underwent PCI; 34.8% underwent CABG.
Early (30-day) clinical outcomes
We examined the impact of revascularization modality on the outcomes reported in the original FREEDOM trial, namely MACCE, as well as each of the components of MACCE. The outcomes are reported as counts and crude event rates in Table 3. The interaction between acuity (ACS, SIHD) and treatment (PCI, CABG) was significant (p < 0.01). Therefore the early event rates for MACCE and its components are presented by acuity. The rates were lower in patients undergoing CABG compared with PCI for MACCE, MI, and death, but these differences were statistically significant only for MACCE and MI. Conversely, stroke rates were significantly higher in the CABG group.
When expressed as ORs, CABG-treated patients had significantly lower odds of MACCE and nonfatal MI compared with patients who underwent PCI. Although the odds of all-cause mortality were also lower in the CABG-treated patients, this did not reach statistical significance. After multivariable adjustment for baseline differences, the odds of MACCE in the CABG group remained significantly lower at 0.60 (95% CI: 0.43 to 0.84). The IPTW method–based odds of MACCE were consistent with the results of the multivariable adjusted model (OR: 0.62; 95% CI: 0.46 to, 0.84). The ORs for each of the component outcomes favored CABG, as shown in Figure 2, except for stroke.
Impact of presentation with ACS on early (30-day) clinical outcomes
The impact of revascularization strategy on MACCE varied significantly by acuity of presentation (ACS vs. SIHD; pinteraction <0.01) and also by disease severity (3-vessel disease vs. 2-vessel disease; pinteraction = 0.05). Among ACS patients undergoing CABG compared with PCI, the adjusted OR for MACCE strongly favored CABG (adjusted OR: 0.49; 95% CI: 0.34 to 0.71), whereas among patients with SIHD, the odds of MACCE did not vary significantly on the basis of revascularization strategy (OR: 1.46; 95% CI: 0.71 to 3.01). The IPTW-based model corroborated the findings from the multivariable adjusted analysis, with a significant pinteraction of 0.02. Among ACS patients, the OR for MACCE favored CABG (OR: 0.52; 95% CI: 0.37 to 0.73), whereas among SIHD patients, MACCE was not affected by revascularization strategy (OR: 1.16; 95% CI: 0.63 to 2.11).
Similarly, among patients with triple-vessel disease who were undergoing CABG compared with PCI, the adjusted OR for MACCE favored CABG (adjusted OR: 0.47; 95% CI: 0.31 to 0.71), whereas the advantage of CABG over PCI in patients with 2-vessel disease was not significant (adjusted OR: 0.88; 95% CI: 0.54 to 1.44).
Late (31-day to 5-year) clinical outcomes
The late Kaplan-Meier event rates for MACCE and its component outcomes are presented in Table 4. The interaction between acuity (ACS, SIHD) and treatment (PCI, CABG) was not significant (p = 0.39); nevertheless, for consistency with the short-term outcomes, we present the results stratified by acuity. MACCE and all the component outcomes, including repeat revascularization, were all lower in the CABG-treated patients. These differences were all statistically significant, except for stroke. When expressed as HRs, CABG-treated patients had a significantly lower risk of MACCE (adjusted HR: 0.63; 95% CI: 0.53 to 0.74; IPTW adjusted HR: 0.69; 95% CI: 0.59 to 0.81), MACCE(r) (HR: 0.40; 95% CI: 0.35 to 0.46), and the component outcomes of all-cause mortality, MI, stroke, and repeat revascularization, compared with PCI-treated patients (Figure 3). Post-30-days, the all-cause mortality rate was lower by 52% (HR: 0.48; 95% CI: 0.39 to 0.59), MI by 60% (HR: 0.40; 95% CI: 0.31 to 0.51), stroke by 20% (HR: 0.80; 95% CI: 0.56 to 1.15), and repeat revascularization by 72% (HR: 0.28; 95% CI: 0.23 to 0.35) in patients undergoing CABG compared with PCI. The results from the IPTW-based model for the component outcomes were consistent with the results reported earlier and were only minimally attenuated. The cumulative incidence curves are plotted in Figure 4, thereby demonstrating the higher risk for all outcomes in the PCI-treated patients and continuing divergence of the curves over time. The median follow-up time was 3.3 years (interquartile range [IQR]: 1.8 to 4.9 years).
The late benefit of CABG over PCI no longer varied by acuity of presentation. In patients presenting with ACS or SIHD, the use of CABG was associated with a significantly lower risk of MACCE (HR: 0.67; 95% CI: 0.55 to 0.81; and HR: 0.55; 95% CI: 0.40 to 0.74, respectively; pinteraction = 0.28). Again, the IPTW-based model corroborated these results (HR for MACCE in ACS patients: 0.74; 95% CI: 0.62 to 0.89; and HR for MACCE in SIHD patients: 0.56; 95% CI: 0.41 to 0.77). Figure 5 illustrates the differences between CABG- and PCI-treated patients before and after applying PS-IPTW). Importantly, after applying IPTW, all between-group standardized differences were <10.
The study time period encompasses a marked change from bare-metal stents to DESs and among DESs from first-generation to second-generation stents for PCI. We conducted further sensitivity analyses to determine whether the overall advantage of CABG is attenuated with an increase in PCI with DES use from 28% to 95% and/or the introduction of second-generation DESs for PCI in 2010, with a subsequent rise in use to 93% (Online Figure 1).
The results suggest that the use of DESs had little impact on the results reported for CABG versus PCI, both in the early follow-up period and in the late follow-up period. Among ACS patients, CABG versus all PCI demonstrated an advantage of CABG (OR: 0.49; 95% CI: 0.34 to 0.71); this advantage was essentially unchanged when restricted to DESs only (OR: 0.51; 95% CI: 0.35 to 0.77). Among SIHD patients, there was no significant difference between CABG versus all PCI on MACCE (OR: 1.46; 95% CI: 0.71 to 3.01), and this remained unchanged when restricted to CABG versus DESs (OR: 1.46; 95% CI: 0.67 to 3.19). Similarly, in the late follow-up, the advantage of CABG over all PCIs (HR: 0.63; 95% CI: 0.53 to 0.74) was essentially unchanged when CABG was compared with DESs (HR: 0.74; 95% CI: 0.62 to 0.89).
With respect to second-generation DES stent use, we examined the impact of year of procedure on MACCE by constructing an interaction term—procedure year by procedure type—in which procedure type was defined as CABG or DESs or bare-metal stents, and procedure year is a surrogate for type of DES (first- or second-generation stent). Although we have aggregate information on type of DESs, we do not have these data at the patient level. Procedure year allows us to determine whether the low use of second-generation DESs before 2010, followed by a rapid uptake reaching 93% by 2014, attenuated the advantage of CABG over PCI. There was no further attenuation of the advantage of CABG over PCI on outcomes, in the early or late follow-up period, when procedure year, as a proxy for second-generation DES use, was included. For further details, see Online Table 1.
In addition, to assess the extent to which the treatment estimate could be affected by selection bias from unmeasured confounders, we performed an IV analysis. The by-year and by-hospital proportion of CABG procedures, on the basis of the hospital and year of diagnostic catheterization, reflects the treatment selection tendency (CABG vs. PCI) for a given year at a given hospital and was used as our instrument in IV analysis. The assessment of the IV indicates minimal bias in treatment effect as a result of potential unobserved confounders (r = 0.04; p = 0.789), and the strength of our IV, on the basis of a partial F test statistic of 201.2 (p < 0.001) is substantially greater than the conventional threshold of 10. For the entire cohort, the OR for CABG versus PCI on 30-day MACCE was 0.79 (95% CI: 0.34 to 1.83). When restricted to the years 2011 through 2014, when second-generation DES use was >80%, the OR was 0.52 (95% CI: 0.19 to 1.44).
In our large, observational, real-world study of CABG versus PCI, among patients with DM and MV-CAD, MACCE outcomes were lower in those patients who underwent CABG (Central Illustration). During long-term follow-up, all individual components of MACCE were lower in patients undergoing CABG, notably a statistically significant 52% relative risk reduction in all-cause mortality.
The controversy about the most efficacious mode of revascularization in SIHD patients with DM and MV-CAD was clarified by the results of the FREEDOM trial (3) and in subsequent meta-analysis of randomized trials (4). In FREEDOM, among patients receiving aggressive evidence-driven medical therapy, the 5-year primary composite outcome occurred in 18.7% of the CABG group and in 26.6% of the PCI-DES group (p = 0.005) (3). Importantly, the benefit of CABG was driven by highly significant absolute reduction in MI (7.9%; p < 0.001), but with a higher risk of stroke in the CABG group (5.2% vs. 2.4%; p = 0.03). Our real-world observational analysis complements the effectiveness of CABG over PCI in the randomized clinical trial findings in SIHD patients. When compared with FREEDOM, our cohort was older and had more comorbidities such as left ventricular impairment (LVEF < 40%) (4.7% vs. 2.5% in FREEDOM) but considerably lower rates of 3-vessel disease (43% vs. 83%). Both the observational and the randomized cohorts were similar in incidence of hypertension and in percentage of female sex (3). Most importantly, FREEDOM excluded patients presenting with ACS within 72 h, whereas our cohort included all-comers.
Registries to better inform clinical trial findings are often conducted concurrently (8). However, a formal FREEDOM registry was not prospectively funded. Within the confines of the Cardiac Services British Columbia registry we set out to overcome this shortfall by using as many of the FREEDOM trial inclusion and exclusion criteria as possible and using provincial administrative data for clinical events. To maximize the reliability of our data and achieve the highest-quality results, we limited the observational cohort to include only those patients undergoing CABG or PCI after October 2007, to reflect contemporary practices (surgical techniques, early invasive angiography, use of DESs, and use of dual antiplatelet therapy) most accurately.
Overall, PCI (60%) was chosen twice as often as CABG in our cohort, similar to what was observed in the BARI (Bypass Angioplasty Revascularization Investigation) registry (percutaneous transluminal coronary angioplasty [PTCA] vs. CABG in patients with MV-CAD) (9). In the BARI registry both the PTCA and CABG cohorts were well matched for clinical characteristics, though anatomically the CABG-treated patients had more 3-vessel diseases and proximal left anterior descending coronary artery involvement. A similar and substantial number in both groups had a diagnosis of ACS. Importantly, to be enrolled in the BARI registry, patients were eligible for both revascularization strategies, which may have not been the case in a real-world clinical database. The BARI registry PTCA-treated patients had better outcomes when compared with the registry patients on the basis of better patient selection. Similarly, with technological advances over the past 2 decades in PCI (delivery systems, stents, and adjunctive pharmacotherapy), there may have been willingness on the part of the interventional cardiologists to undertake more challenging anatomic cases, thereby contributing to the differences we observed in favor of CABG over PCI.
There were substantial clinical and angiographic differences between the PCI and CABG groups in our cohort, with many of the unfavorable clinical characteristics (age, sex, comorbidities, and lower LVEF) being more common in the PCI group. These important differences, combined with higher rates of ACS patients in the PCI group, is entirely consistent with the current practice of generally higher-acuity patients and those declined for CABG undergoing high-risk PCI. We attempted to overcome some of the clinical differences through statistical adjustment of important variables. To minimize differences further, we excluded patients with cardiogenic shock or those requiring mechanical support. Furthermore, if LVEF assessment was not available, these patients were categorized in the highest risk category in our modeling. Conversely, it is also possible that given the high percentage of patients undergoing ad hoc PCI, those patients with the most favorable anatomy and those with lower disease burden may have undergone PCI without consideration of CABG.
Prior studies of CABG versus PCI in patients with DM and MV-CAD in the DES era mainly enrolled SIHD patients (4). In contrast, more than 63% of our cohort consisted of patients with ACS, thus reflecting an interesting practice for DM patients in British Columbia: a higher rate of CABG as the initial revascularization strategy compared with U.S. registries. The determinants to undergo ad hoc PCI are likely multiple and complex and include patients’ and physicians’ preference, upstream use of dual antiplatelet therapy, delayed availability of CABG, anatomy or comorbidities not suited for CABG, and other specific local institutional factors. The ACS patients compared with the SIHD patients in our cohort were slightly older, had more comorbidities, and had similar anatomic variables.
In the past, the perceived increased risk with CABG in ACS compared with non-ACS patients was a rationale for considering PCI over CABG. However, the choice of revascularization is more complex and includes the severity and extent of disease, the extent of ischemia, procedural risks, durability of the results, and completeness of revascularization; many of these elements are not assessed in this population-based registry. At present, only post hoc retrospective data with variable durations of follow-up are available for culprit lesion PCI versus multivessel PCI in ACS patients with non–ST-segment elevation MI (10). Generally, CABG is preferentially selected over multivessel PCI for patients with more extensive disease burden. Our cohort also demonstrated that CABG-treated patients had a significantly higher burden of CAD (p < 0.01), whereas PCI-treated patients were older, more likely to be female, and had comorbidities and lower LVEF (p < 0.01 for all). Currently, there is controversy about the optimal timing of CABG following an ACS. The safe time interval between myocardial injury and CABG was beyond 90 days in the European System for Cardiac Operative Risk Evaluation (EuroSCORE) (11). Early surgery post-myocardial injury can be associated with edema causing poor visualization of the target vessel and arteriotomy site. Furthermore, successful reperfusion of an occluded vessel can result in reperfusion injury (12). Conversely, early CABG may lead to improved left ventricular systolic function and decreased arrhythmias. With improved surgical techniques and periprocedural operative care, there is increasing acceptance to carry out CABG following an ACS. Davierwala et al. (13), in a single-center study of 758 patients, showed that patients with CABG performed within 24 h of a non–ST-segment elevation MI had in-hospital mortality rates and long-term outcomes similar to those having CABG after 3-days. Among the patients undergoing CABG in our cohort, the procedure was deemed necessary during the index hospitalization by the treating team. The median time from cardiac catheterization to CABG in the overall cohort was 14.9 days (IQR: 6.7 to 65.0 days) compared with the ACS group of 7.8 days (IQR: 4.8 to 13.9 days), thus reflecting 2 distinct patterns of practice.
The early and late benefits seen with CABG in our cohort of mainly ACS patients raise the possibility that in the current era, further gains may be made by moving beyond ad hoc PCI as the default procedure in diabetic patients with MV-CAD. Our results need to be validated by other large, population-based registries, and a randomized controlled trial of CABG versus PCI in the ACS population is needed to inform practice guidelines.
In a network meta-analysis of 100 trials and more than 90,000 patients, in which revascularization was compared with medical therapy for SIHD, Windecker et al. (14) found improved survival compared with medical therapy with CABG as well as with newer-generation stents, but not older PCI technologies. Indeed, the findings were similar even when including patients with recent ACS. Similarly, Bangalore et al. (15) reported indirect comparisons of patients undergoing CABG with PCI specifically with DM and demonstrated similar mortality rates with either strategy. Such analyses continue to raise the important question whether advancing PCI technology will yield a different result from what has been seen thus far in comparative revascularization trials; the FAME 3 (A Comparison of Fractional Flow Reserve-Guided Percutaneous Coronary Intervention and Coronary Artery Bypass Graft Surgery in Patients With Multivessel Coronary Artery Disease) trial seeks to address this hypothesis in patients with 3-vessel disease, by using a newer-generation stent platform in concert with fractional flow reserve guidance (16). To date, however, such analyses are limited by varying extents of 3-vessel CAD, variable inclusion of ACS patients, and varying degrees of follow-up. Also necessary to consider is that the FREEDOM trial was largely a trial of 3-vessel disease (83% of patients), whereas in our study 42% to 43% of the patients 3-vessel disease. The fact that our findings were limited to those patients with 3-vessel disease suggests that the extent of CAD is an important discriminator of benefit from higher-order therapy. Taken in total, these hypothesis-generating findings reinforce the critical need for well-powered randomized trials, most notably in patients with DM.
This study has the inherent limitations of an observational registry, with only selected baseline characteristics gathered, limited data on background therapy, and uncertainty on adherence to guideline-directed management of risk factors. Strict protocol-driven follow-up is unavailable outside the confines of a clinical trial, and as such it is possible that our analysis may truly underestimate the number of events. However, because the nature of the events of interest in the FREEDOM trial would have resulted in hospitalization, it is unlikely that we have missed significant events because each patient has a unique personal health number that allows us to link to the province-wide hospitalization database. Similarly, linkage with the provincial vital statistics database ensures that we capture all deaths, and linkage with the Cardiac Services British Columbia registry ensures that we capture all repeat revascularization procedures.
There are also some analytical limitations with missing variables, notably details on kidney function. However, the number of patients with severe renal failure in the range of renal replacement therapy was similar in both groups, and it was small. Therefore, the applicability of our findings specifically in patients with moderate to severe kidney function cannot be substantiated. We also have limited anatomic and procedural data and therefore cannot comment on the complexity of disease SYNTAX (Synergy Between PCI With Taxus and Cardiac Surgery) scores and completeness of revascularization and on the influence of these variables on outcomes. Finally, our patient population consisted primarily of patients undergoing ad hoc PCI compared with patients having CABG after a period of stabilization. Therefore, it is possible that sicker patients were being selected for PCI for their acuity. Moreover, there may have been a survival bias because of the delay between cardiac catheterization and CABG, with the frailer and older patients dying before their CABG procedures (17). Furthermore, because we lack patient-level data on type of DESs, we could not provide a robust comparison between CABG and second-generation DESs.
Despite multivariable adjustment and PS-IPTW based modeling, and the robustness of the findings when using both these methods, it is still possible that important, unmeasured confounders were not accounted for. Observational studies can address only measured confounders. If an unmeasured confounder both has a strong effect on our outcomes of interest and is differentially distributed by mode of revascularization, this could bias the results. To assess this possibility, we carried out an IV analysis, which indicated minimal bias in treatment effects as a result of unobserved confounders. Although the IV analysis results were not statistically significant, this is not surprising because IV analyses tend to generate wider 95% CIs, even when the IV is exogenous and relatively strong. However, the consistency of the study findings with the findings of the randomized control trials, the comprehensiveness of the clinical data available for multivariable adjustment and PS-IPTW–based modeling, and the inclusion and follow-up of all patients undergoing revascularization during the study time period strengthen the validity of the study findings.
In a large, contemporary, and validated database, we provide robust evidence that CABG was associated with better outcomes compared with PCI for MV-CAD at a population level. Importantly, these benefits are driven by a marked 37% superiority of CABG (over PCI) in terms of MACCE outcomes and a 52% reduction in all-cause mortality in the long term. The long-term cardiovascular benefits of CABG appear to be present in patients with and without ACS. For patients with DM who present with MV-CAD, these data provide extrinsic validation of the randomized trials for patients with SIHD and represent a call to action for a large definitive randomized trial of patients presenting with ACS.
COMPETENCY IN PATIENT CARE: In patients with DM and MV-CAD stabilized following an ACS, as for those patients with SIHD, a “heart team” approach is recommended to individualize care, but CABG is generally the preferred method of revascularization.
TRANSLATIONAL OUTLOOK: Patients with DM and MV-CAD who survive an ACS have been underrepresented in clinical trials, and further prospective studies are needed to define optimum revascularization strategies.
The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
Sripal Bangalore, MD, MHA, served as Guest Editor for this paper.
Presented in part at the Annual Scientific Sessions of the American Heart Association, Orlando, Florida, November 2015.
- Abbreviations and Acronyms
- acute coronary syndrome(s)
- coronary artery bypass grafting
- coronary artery disease
- confidence interval
- drug-eluting stent
- diabetes mellitus
- hazard ratio
- interquartile range
- instrumental variable
- left ventricular ejection fraction
- major adverse cardiac or cerebrovascular event(s)
- composite of major adverse cardiac or cerebrovascular event and repeat revascularization
- myocardial infarction
- multivessel coronary artery disease
- odds ratio
- percutaneous coronary intervention
- stable ischemic heart disease
- Received October 21, 2016.
- Revision received September 29, 2017.
- Accepted October 10, 2017.
- 2017 American College of Cardiology Foundation
- Amsterdam E.A.,
- Wenger N.K.,
- Brindis R.G.,
- et al.
- McCormick N.,
- Lacaille D.,
- Bhole V.,
- Avina-Zubieta J.A.
- Brookhart M.A.,
- Wyss R.,
- Layton J.B.,
- Sturmer I.
- Hochman J.S.,
- Buller C.E.,
- Sleeper L.A.,
- et al.
- Feit F.,
- Brooks M.M.,
- Sopko G.,
- et al.
- Davierwala P.M.,
- Verevkin A.,
- Leontyev S.,
- Misfeld M.,
- Borger M.A.,
- Mohr F.W.
- Windecker S.,
- Stortecky S.,
- Stefanini G.G.,
- et al.
- Bangalore S.,
- Toklu B.,
- Feit F.
- ↵Fearon W. A Comparison of Fractional Flow Reserve-Guided Percutaneous Coronary Intervention and Coronary Artery Bypass Graft Surgery in Patients With Multivessel Coronary Artery Disease (FAME 3). Stanford University. Trial identifier: NCT02100722. Available at: clinicaltrials.gov. Accessed October 23, 2017.