Revascularization Trends in Patients With Diabetes Mellitus and Multivessel Coronary Artery Disease Presenting With Non–ST Elevation Myocardial Infarction

Introduction

Diabetes mellitus (DM) is a rapidly expanding global pandemic, affecting over 350 million people worldwide.^1,2 DM is also one of the most important risk factors for coronary artery disease (CAD), and 40% of patients with acute coronary syndrome (ACS) have DM.^3–5 Among patients presenting with an ACS, patients with DM are at least 80% more likely than patients without DM to have multivessel CAD on angiography.^6,7

Current American College of Cardiology Foundation/American Heart Association guidelines recommend coronary artery bypass graft (CABG) over percutaneous coronary intervention (PCI) as the preferred revascularization strategy in patients with multivessel CAD and DM (Class IIA).⁸ These recommendations are based largely on randomized control trials enrolling patients with stable CAD.^9–13 However, in a subgroup analysis of the Future Revascularization Evaluation in Patients with Diabetes Mellitus: Optimal Management of Multivessel Disease (FREEDOM) trial, in which approximately one third of the cohort (n=583) was enrolled after an ACS event, the superiority of CABG in the ACS subset was comparable to the overall trial population.¹¹

Despite the guideline recommendations favoring CABG, previous studies have identified an implementation gap between the available evidence base and routine medical practice in patients with DM, with at least 1 in 4 eligible patients with multivessel CAD receiving PCI instead of CABG.¹⁴ In addition, revascularization decisions in patients presenting with non–ST-segment–elevation myocardial infarction (NSTEMI) may be different than patients with stable CAD. A previous study from the Can Rapid Risk Stratification of Unstable Angina Patients Suppress Adverse Outcomes With Early Implementation of the ACC/AHA Guidelines (CRUSADE) Quality Improvement Initiative reported greater use of PCI as compared with CABG among NSTEMI patients with multivessel disease in the early 2000s.¹⁵ However, contemporary trends in revascularization among the high-risk subset of participants with multivessel CAD and type 2 DM are not known. Accordingly, we used the National Cardiovascular Data Registry (NCDR) Acute Coronary Treatment and Intervention Outcomes Network Registry–Get with the Guidelines (ACTION Registry–GWTG) database to study temporal trends and hospital-level variation in performance of CABG and PCI among DM patients with NSTEMI and angiography, demonstrating multivessel CAD. We also sought to determine the major clinical factors associated with use of PCI compared with CABG.

Methods

Data Source

The present study was performed using data obtained from the NCDR ACTION Registry-GWTG, a large national quality improvement registry that focuses exclusively on patients presenting with an acute myocardial infarction. Details of the ACTION Registry-GWTG have been previously described.¹⁶ The registry uses trained hospital personnel from >1000 participating centers across the United States to collect data from medical records using a standardized data set with written definitions (https://www.ncdr.com/webncdr/action/home/datacollection). Abstracted data include patient demographics, clinical information, medical therapies, use and timing of cardiac procedures, and in-hospital outcomes. The registry has requirements in place to ensure uniform data entry and transmission and is subject to data quality checks (www.ncdr.com). Because all data are abstracted retrospectively and anonymously without unique patient identifiers, institutional review boards by policy waive the need for informed consent.

Study Population

For the present study, we included DM patients who presented with NSTEMI between July 2008 and December 2014 and had angiography demonstrating multivessel CAD (n=74 941). Major exclusion criteria included transfer out of the presenting hospital for further management (n=3707), prior history of revascularization (n=38 040), hemodynamic instability or cardiac arrest at presentation (n=1041), and presentation to centers lacking CABG and PCI capabilities (n=2384; Figure 1). The final study population included 29 769 patients from 539 hospitals. Prevalent DM was defined as a documented medical history of DM. Multivessel obstructive disease was defined as ≥50% stenosis in left main artery or ≥70% stenosis in ≥2 separate epicardial coronary artery territories. This definition of multivessel obstructive disease is consistent with that used in the FREEDOM trial.¹¹

• Download figure
• Download PowerPoint

Statistical Analysis

Patients were categorized into 3 groups: CABG versus PCI versus no-revascularization. Approximately 2% of patients received both CABG and PCI and were classified as being treated with CABG. Baseline characteristics and in-hospital treatment and discharge medications were compared across the 3 groups using the χ² test for categorical variables and Kruskal–Wallis test for continuous variables. The Kruskal–Wallis test was used rather than other procedures, such as the ANOVA test, for means because the test is a legitimate test in the presence of normality as well as non-normality. Consistency of a single test across continuous variables was preferred. The differences in proportional use of different management strategies (CABG versus PCI versus no revascularization) in the primary analysis cohort during each quarter (Q) of the analysis time period (Q3 2008 to Q4 2014) were presented for each quarter, and the test for a linear trend among each treatment strategy was obtained by modeling quarter of patient hospital arrival as an ordinal independent variable using logistic regression. Hierarchical logistic regression modeling with hospital-specific random intercepts was used to test whether the CABG and PCI rates were different between hospitals. This strategy describes the true signal variation in the proportion of patients who receive CABG or PCI while removing the effect of random sample variation.¹⁷ Because the range in rate would vary greatly between hospitals with too few patients, this analysis was restricted to hospitals with at least 25 patients over the study period (2008–2014) in the calculation of each rate. This threshold has been previously used in other studies of hospital variation.^18,19 The raw proportion and corresponding 95% confidence intervals for each hospital are displayed on each hospital-level variation figure. Subgroup analyses were performed to evaluate the temporal trends and hospital-specific variability in use of these revascularization strategies among participants stratified by number of vessels involved (2-vessel or 3-vessel CAD) and presence or absence of proximal left anterior descending artery (pLAD) stenosis ≥70%.

Factors associated with use of CABG versus PCI among patients with DM who underwent revascularization were identified by constructing a hierarchical logistic regression model where hospital was included as a random effect to account for variability between hospitals for the binary outcome CABG versus PCI, using all the clinical and demographic characteristics, as detailed in Table in the Data Supplement. Backward selection was used for the model with a retention P value criterion of <0.05 for the final multivariable model. Additionally, we repeated the modeling process using stepwise variable selection with an alpha to enter of 0.15 and an alpha to remove of 0.05, and the final model was exactly the same. All continuous variables were evaluated for nonlinearity with the outcome, and linear splines were used for those that did not meet the linear relationship criteria. All candidate variables had a missing rate of <2%. Variables with missing data were imputed to the median of the nonmissing values for continuous variables or to the most frequently occurring group for categorical variables. Body mass index was imputed to the sex-specific median for missing values. SAS version 9.4 (SAS Institute, Cary, NC) was used to perform all analyses in the present study. A 2-sided P value of 0.05 was considered significant. All confidence intervals were calculated at the 95% level.

Results

Baseline Characteristics of Study Population

A total of 29 769 patients with DM from 539 participating centers that presented with NSTEMI and had coronary angiography revealing multivessel CAD were included for analysis. Of these, 36.4% were treated with CABG, 46.2% were treated with PCI (77.2% with at least 1 drug-eluting stent), and 17.3% were treated without revascularization. Baseline characteristics of the study population stratified by treatment strategy (CABG versus PCI versus no-revascularization) are shown in Table 1. Compared with patients treated with PCI, those treated with CABG were more likely to be men, white, and have lower peak troponin levels; they were more likely to have electrocardiographic changes and heart failure at presentation, more likely to have left main, pLAD, and 3-vessel disease on coronary angiography, and also less likely to receive early treatment with adenosine diphosphate receptor antagonist on presentation. Patients in the no-revascularization group were older, more likely to be women, with a higher burden of CAD risk factors and comorbidities as compared with patients who underwent CABG or PCI. Minimal differences were noted in patient characteristics over time (data not shown).

Temporal Trends in Use of Revascularization Strategies

The proportion of patients revascularized increased modestly over time from 81.1% in 2008-Q3 to 83.6% in 2014-Q4 (P_trend<0.0001) attributable entirely to increasing use of PCI (45% in 2008-Q3 to 48.9% in 2014-Q4; P_trend=0.0002), with no change in the use of CABG (36.1% in 2008-Q3 to 34.7 in 2014 Q4; P_trend=0.88; Figure 2).

• Download figure
• Download PowerPoint

Analyses were performed after stratification by the angiographic severity of CAD. The relative proportion of patients undergoing CABG versus PCI versus medical management varied significantly based on the severity of CAD with CABG rates of 14.9% among those with low-risk 2-vessel CAD without pLAD involvement to 58.0% among patients with high-risk 3-vessel CAD with pLAD involvement. (Figure 3) PCI rates increased significantly over time among participants with 3-vessel CAD without pLAD involvement with no significant changes in rates of CABG or PCI among other angiographic subsets (Figure I in the Data Supplement).

• Download figure
• Download PowerPoint

Hospital-Specific Variability in Revascularization Strategies

In the primary analysis cohort of 312 hospitals treating at least 25 patients throughout the study period, the hospital-level use of CABG and PCI varied widely (CABG range =0%–78%; PCI range =22%–100%; P value <0.0001 for both; Figure 4). Similarly, we observed significant hospital-specific variability in use of CABG and PCI among subgroups defined by the severity of CAD (Figure II in the Data Supplement).

• Download figure
• Download PowerPoint

Factors Associated With Use of CABG Versus PCI

Significant factors associated with CABG versus PCI are presented in Table 2. CABG versus PCI is a binary outcome in which logistic regression modeling is appropriate. Also, continuous variables that were nonlinear with respect to the outcome were fit with linear spline variables. For example, age was not linear with respect to the outcome and was fit with linear piece-wise splines at 72 years. The presence of left main artery or pLAD stenosis, signs of heart failure and elevated heart rate on presentation, mild to moderate left ventricular dysfunction, and increasing age (in patients <72 years) were associated with a greater likelihood of CABG as compared with PCI. In contrast, early use of adenosine diphosphate receptor antagonists, such as clopidogrel or prasugrel on presentation, female sex, history of smoking, prior myocardial infarction, prior heart failure, current dialysis use, increasing age (in patients >72), higher body mass index (in overweight and obese patients), and Medicare insurance coverage (versus HMO or private insurance) were associated with a lower likelihood of CABG (Table 2).

Discussion

The principal finding of this study is that only one third of DM patients with NSTEMI and multivessel CAD undergo CABG during index hospitalization. PCI use has increased modestly over time in such patients with no change in the proportion undergoing CABG. The differential use of CABG versus PCI among these participants seems to be related both to the anatomic severity of underlying CAD as well as patient factors, such as age, sex, and body mass index. Additionally, there are marked variations in the proportional use of CABG and PCI among participating hospitals, which persist even among angiographic subgroups, a finding that suggests that local practice patterns play an important role in selection of CABG versus PCI in NSTEMI patients with DM and multivessel CAD. To our knowledge, this is one of the largest and most contemporary evaluations of trends in revascularization among patients with DM and multivessel coronary disease presenting with NSTEMI.

Following the results of the Bypass Angioplasty Revascularization Investigation (BARI) trial demonstrating superior survival with CABG over balloon angioplasty in the subset of patients with DM, the National Heart Lung and Blood Institute issued a Clinical Alert in 1995 in favor of CABG over PCI for patients with DM and multivessel CAD. However, the proportion of patients with DM and multivessel CAD undergoing PCI remained unchanged after the clinical alert (28.6% before versus 26.8% after the Clinical Alert).²⁰ Since then, other randomized clinical trials and a recent meta-analysis have all consistently demonstrated significantly improved clinical outcomes, including mortality and repeat revascularization, with CABG over PCI in patients with DM and multivessel CAD.^10–12,21 However, most of these studies predominantly included patients with stable disease, and the optimal revascularization strategy among these patients in the setting of an ACS presentation is not known. Furthermore, these trials compared CABG against PCI with bare metal stents and first-generation drug-eluting stent (paclitaxel or sirolimus-eluting stents), and findings from recent studies suggest that the superiority of CABG over PCI may be less clear with the use of newer generation drug-eluting stent.^22–24

The results from our present analysis provide an insight into the contemporary patterns of revascularization among patients with DM and multivessel CAD in the setting of NSTEMI. We observed greater utilization of PCI versus CABG in this patient population throughout the study period. Furthermore, consistent with previous cohort studies,^15,22,25 we observed that the choice of optimal revascularization strategy in this patient population is particularly influenced by the severity of underlying CAD. Among high-risk patients (3-vessel CAD and pLAD involvement), CABG has been the revascularization strategy of choice throughout the study period. Similarly, PCI is the preferred revascularization strategy among lower-risk patients (2-vessel CAD without pLAD involvement) with no significant change over time. Furthermore, the proportional site-specific variability in use of CABG in the high-risk subgroup (3-vessel CAD with pLAD) and PCI in the low-risk subgroup (2-vessel CAD without pLAD) is lower when compared with the overall population (CABG range of 45%–96% in high-risk group versus 0%–78% in the overall cohort; PCI range of 51%–98% in the low-risk subgroup versus 22%–100% in the overall population). This anatomic disease severity–driven revascularization practice pattern observed in our study is consistent with the emerging evidence base as noted in the Synergy between PCI with Taxus and Cardiac Surgery (SYNTAX) trial.¹² Other patient characterstics most strongly associated with greater likelihood for CABG were presence of significant ST changes, signs of heart failure, mild to moderately depressed ejection fraction, and tachycardia at presentation. In contrast, high surgical risk characterstics, such as older age, female sex, obesity, presence of end-stage renal disease, and prior history of myocardial infarction and congestive heart failure, were associated with lower likelihood of CABG. This is also consistent with findings reported from previous cohort studies.^15,22,25

Several other factors may also influence the choice of revascularization strategy in favor of PCI over CABG in this patient population. First, revascularization decisions in usual clinical practice may be different in patients with NSTEMI than those with stable CAD. For example, although patients with cardiac arrest at presentation and hemodynamic instability were excluded in the present analyses, other features of NSTEMI clinical presentation may require consideration for urgent percutaneous intervention, such as ongoing symptoms, high Global Registry of Acute Coronary Events (GRACE) score at presentation, and evidence of thrombus with impaired Thrombolysis In Myocardial Infarction flow on initial angiography, which is estimated to occur in ≤25% of patients with NSTEMI.²⁶ Second, nearly 50% of patients presenting with NSTEMI routinely receive pretreatment with adenosine diphosphate receptor antagonists on presentation.²⁷ This confounds consideration for early CABG, making PCI a more favorable option to affect prompt revascularization, as also observed in our study. Third, other issues such as operative mortality and morbidity considerations and longer length of stay with CABG can also be important considerations.¹³ Finally, most of the randomized controlled trials comparing PCI with CABG enrolled highly selected stable CAD patients who were considered suitable candidates for either procedure and who were free from major comorbidities that would be expected to preclude long-term survival.

The increasing use of PCI observed in the present study adds to a growing body of literature on PCI in patients with multivessel CAD.^15,28,29 A previous study from the NCDR registry demonstrated a temporal increase in use of PCI in patients with multivessel CAD and Class I indication(s) for surgical revascularization (73% with left main or multivessel CAD; 41% with ACS).²⁸ Similarly, in the CRUSADE quality improvement initiative, investigators reported that among patients with NSTEMI and 3-vessel disease, the rate of PCI has increased (51.1%–60.1%) while that of CABG decreased over time between 2001 and 2004.¹⁵ Our findings add significantly to these studies by evaluating a much higher risk subgroup of the multivessel CAD patient population with underlying DM and an ACS presentation.

We observed significant hospital-level variation in the proportional use of CABG and PCI across participating centers. This marked variation likely indicates a lack of clinical consensus among physicians regarding the optimal revascularization strategy for these patients, rather than a lack of awareness regarding the evidence base, as previously noted.¹⁴ Interventional cardiologists may interpret the available evidence demonstrating superiority of CABG over PCI in the stable CAD population differently in the setting of NSTEMI. Although some may generalize these results to patients with a more acute presentation, others may choose to be more restrictive in their approach, considering the available evidence not applicable in the setting of NSTEMI. The provider’s preferences may also bias the patients to align toward less invasive option.³⁰ Furthermore, differences in local expertise of interventional cardiologists and cardiothoracic surgeons and systems of perioperative care could also influence the decision-making process.²⁹ Future studies focused on risk perception of patients, and operator-level practice patterns are needed to better understand the drivers of the observed variability in revascularization practice in this patient population.

Our study has several limitations. First, detailed data on angiographic characteristics such as SYNTAX score and presence of thrombus that may influence revascularization decisions were not available. Second, data on important comorbidities, such as malignancies, lack of good coronary bypass targets, and other relative or absolute contraindications to CABG that could influence the revascularization decision, were not available. As a result, real-world complex clinical decision-making could not be elucidated from the database. Third, only the revascularization procedures performed during the index NSTEMI admission were captured in the present study, and we cannot determine what proportion of individuals receiving medical management in the hospital was referred for outpatient CABG or PCI. Next, participation in the registry is voluntary for hospitals. As a result, certain geographical areas are under-represented, and center-level analysis may not be representative of the broader group of hospital in the Unites States. Finally, our study captures data only 2 year after the FREEDOM trial, and the observed findings could be related to an implementation lag. Future studies examining these patterns would be needed to determine whether FREEDOM trial findings significantly influenced revascularization practices among these patients.

In conclusion, only a third of type 2 DM patients with NSTEMI and multivessel CAD undergo CABG during index admission, contrasted with almost half of such patients undergoing PCI. Over the study period, the proportional use of CABG has been stable, whereas that of PCI has been slowly increasing, with practice patterns significantly influenced by the anatomic disease severity. Significant hospital-level variability observed in the proportional use of CABG versus PCI in this study also highlights the continued clinical uncertainty with regard to optimal revascularization strategies in this patient population among practitioners. Future studies are needed to better understand the decision-making process for revascularization strategies among these high-risk patients.

Footnotes