Author + information
- Received May 30, 2018
- Revision received September 19, 2018
- Accepted October 30, 2018
- Published online February 4, 2019.
- Andrew Czarnecki, MD, MSca,b,c,∗ (, )
- Feng Qiu, MScb,
- Gabby Elbaz-Greener, MD, MHAa,
- Eric A. Cohen, MDa,c,
- Dennis T. Ko, MD, MSca,b,c,
- Idan Roifman, MD, MSca,b,c and
- Harindra C. Wijeysundera, MD, PhDa,b,c
- aSchulich Heart Centre, Division of Cardiology, Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada
- bInstitute of Clinical Evaluative Sciences, Toronto, Ontario, Canada
- cDepartment of Medicine, University of Toronto, Toronto, Ontario, Canada
- ↵∗Address for correspondence:
Dr. Andrew Czarnecki, Sunnybrook Health Sciences Centre, D-375, 2075 Bayview Avenue, Toronto, Ontario M4N 3M5, Canada.
Objectives The aims of this study were to assess variation in revascularization of asymptomatic patients with stable ischemic heart disease, identify the predictors of variation, and determine if it was associated with clinical outcomes.
Background Management of stable ischemic heart disease in asymptomatic patients with obstructive coronary artery disease is controversial, potentially leading to practice variation.
Methods A retrospective observational cohort study was performed using population-based data from Ontario, Canada, in patients with asymptomatic stable ischemic heart disease and obstructive coronary artery disease. The cohort was divided on the basis of treatment strategy: revascularization or medical therapy. Hospitals were allocated into tertiles of their revascularization ratio. Outcomes included death and nonfatal myocardial infarction. Hierarchical logistic regression was used to assess the predictors of revascularization, with median odds ratios used to quantify variation. Proportional hazards models were used to determine the association between management strategy and outcomes.
Results The cohort included 9,897 patients, 47% treated with medical therapy and 53% with revascularization. Between hospitals, 2-fold variation existed in the ratio of revascularized to medically treated patients. However, the variation across hospitals was not explained by patient, physician, or hospital factors (median odds ratio in null model: 1.25; median odds ratio in full model: 1.31). Revascularization was associated with a hazard ratio of 0.81 (95% confidence interval: 0.69 to 0.96) for death and a hazard ratio of 0.58 (95% confidence interval: 0.46 to 0.73) for myocardial infarction, with this benefit consistent across tertiles of revascularization ratio.
Conclusions Wide variation was observed in revascularization practice that was not explained by known factors. Despite this variation, a clinical benefit was observed with revascularization that was consistent across hospitals.
Coronary artery disease is the leading cause of death in the United States (1). The mainstays of treatment include medical therapy and revascularization. Although the role of revascularization in the setting of acute coronary syndrome is well established, there is continued controversy about indications for revascularization in stable ischemic heart disease (SIHD) (2). Landmark randomized control trials, including COURAGE (Clinical Outcomes Utilizing Revascularization and Aggressive Drug Evaluation) and BARI 2D (Bypass Angioplasty Revascularization Investigation 2 Diabetes), found no difference in clinical outcomes such as mortality when comparing percutaneous coronary intervention (PCI) versus optimal medical therapy in patients with SIHD (3,4). However, critics of these trials have suggested that they were limited by restricted enrollment, specifically that they excluded those with high-risk disease, thereby reducing generalizability. In contrast, a recent meta-analysis and numerous retrospective studies with broader patient populations have demonstrated a potential benefit with revascularization; however, these studies were limited by their observational nature (5–8). Given this uncertainty regarding “hard” outcomes, clinical decisions have often been dictated by the presence and severity of symptoms.
To assist clinicians, appropriate use criteria for revascularization in patients with SIHD have been published that clearly distinguish between patients with and without symptoms (9). Although revascularization in asymptomatic patients with high-risk disease (such as left main stenosis and multivessel disease) is considered “appropriate,” the majority of indications are uncertain, predominantly because of the paucity of previous studies in asymptomatic patients. This uncertainty may manifest as practice variation. Accordingly, our goal was to address this gap in knowledge by assessing the variation in revascularization patterns among asymptomatic patients with SIHD, identifying the predictors of such variation, and, finally, determining if the variation observed was associated with differential clinical outcomes.
We performed a retrospective observational cohort study using administrative population-based databases available at the Institute of Clinical Evaluative Sciences in Ontario, Canada. Ontario is Canada’s largest province, with 13 million people who have universal access to health care services funded by a single payer (the Ministry of Health and Long-Term Care). Ethics approval was obtained from the Sunnybrook Research Ethics Board. On the basis of Ontario’s privacy legislation, the need for individual patient consent was waived.
The primary dataset was a prospective clinical registry maintained by CorHealth Ontario, which collects information regarding all invasive cardiac procedures in Ontario, including cardiac catheterization, PCI, and coronary artery bypass graft (CABG). There are 19 institutions in the province that perform such procedures, and participation in the clinical registry is a mandatory condition for funding. Data are collected regarding patient demographics, symptoms status, clinical indications, comorbid conditions, and procedural data including coronary anatomy. This database has been previously validated and used extensively for research purposes (7,10–12).
Encrypted patient identifiers were used to link this clinical registry to other administrative datasets. The Canadian Health Information Discharge Abstract Database was used to establish patient comorbidities and hospital admissions. The Ontario Drug Benefit Database was used to capture medication use in patients >65 years of age. The Registered Persons Database was used to ascertain vital status. Johns Hopkins ACG System Version 10 was used to identify frailty on the basis of a series of proprietary diagnostic codes associated with frailty-defining conditions.
Our study cohort consisted of patients >20 years of age who underwent coronary angiography for SIHD between October 1, 2008, and October 31, 2013. Only patients with obstructive coronary artery disease were included, defined as a ≥70% stenosis in a major epicardial coronary vessel (except the left main coronary artery, in which a ≥50% stenosis was considered significant). To establish a cohort of asymptomatic patients, only those with Canadian Cardiovascular Society class 0 angina at the time of angiography were included. Patients were excluded if they had acute coronary syndromes (unstable angina, non–ST-segment elevation myocardial infarction, or ST-segment elevation myocardial infarction). We excluded patients who underwent angiography during a hospital admission to ensure that only a truly stable outpatient population was included, consistent with previous research (7,11,13). If more than 1 angiogram was obtained during the study period, the first was used as the index procedure.
Patients were classified on the basis of the initial treatment strategy, which included either revascularization or medical therapy alone. Revascularization was defined as PCI or CABG within 90 days of the index procedure. If revascularization did not occur within this time window, patients were included in the medical therapy group. The 90-day window was chosen on the basis of previous research and current wait times in Ontario (6,10,11,14). We also examined medication use at 90 days after the index angiographic study.
We calculated revascularization ratios for each hospital performing cardiac catheterization on the basis of the proportion of patients who underwent revascularization versus those who received medical therapy only as the initial treatment strategy. Hospitals were then grouped on the basis of tertiles of revascularization ratio and categorized as high, medium, or low.
The primary outcome was all-cause mortality at last follow-up. The secondary outcome was readmission with nonfatal myocardial infarction at last follow-up. Follow-up was available until March 31, 2016.
Baseline characteristics were categorized on the basis of initial treatment strategy, and standard descriptive statistics were used to compare groups. To evaluate the influence of patient-, physician-, and hospital-level factors on the initial treatment strategy, a three-level hierarchal regression model was used. We included a random effect for hospital and a separate random effect for physicians, which was nested within the hospital. Five following sets of models were created: a null model with random effect only; patient factors only; patient and hospital factors; patient and physician factors; and all patient, physician, and hospital factors. Median odds ratios (MORs) were calculated for each model to quantify the magnitude of the effect of clustering in our multilevel model. The MOR is a measure of the variation between different hospitals that is not explained by the modeled risks (15). Factors associated with initial treatment strategy were determined from the model including all variables.
To determine the association among treatment strategy, practice variation, and study outcomes, Cox proportional hazards models were used. The management strategy was modeled as a time-varying covariate to avoid any potential survivorship bias (i.e., all patients were in the medical therapy group until revascularization was actually performed). All patient, physician, and hospital covariates were included. Robust sandwich variance estimators were used to account for clustering at the hospital level. We then examined the association between hospital revascularization ratio and the study outcomes, by using an additive interaction term for high, medium, and low ratios in the model. We also we added an interaction term between revascularization and the presence of normal and abnormal left ventricular (LV) function, to better evaluate the association between LV function and revascularization. In addition, to ensure that patients who underwent PCI and CABG both endured a consistent treatment effect, we performed a sensitivity analysis in which we considered each type of revascularization separately. The proportional hazards models were then repeated for all study outcomes. Finally, to ensure that we adequately accounted for survivorship bias, a landmark sensitivity analysis was performed. In this analysis, we landmarked outcomes at 90 days in lieu of using time-varying covariates, to ensure overall consistency in our outcomes. All analyses were performed using SAS version 9.3 (SAS Institute, Cary, North Carolina). A p value <0.05 was considered to indicate statistical significance.
There were 28,547 procedures performed in patients with SIHD who were asymptomatic during the study period (Figure 1). After excluding patients who were in hospital or <20 years of age (n = 2,410), those with nonobstructive disease (n = 15,026), and those with multiple procedures (n = 329), there were 9,897 patients in the final study cohort. The initial treatment strategy was medical therapy in 4,650 patients (47%) and revascularization in 5,247 patients (21% CABG, 32% PCI). Among patients assigned to the medical therapy group, 164 (3.5%) underwent PCI and 345 (7.4%) underwent CABG within 1 year of index angiography (10.9% of the medical therapy group). Median follow-up for survival was 1,675 days (interquartile range: 1,210 to 2,179 days).
Baseline characteristics according to initial management strategy are shown in Table 1, as is medication use for patients >65 years of age. Patients treated with medical therapy were older than those treated with revascularization, and medical comorbidities were all more prevalent in the medical therapy group. Baseline characteristics subcategorized by type of revascularization (PCI or CABG) and tertile of revascularization ratio are shown in Online Tables 1 and 2.
The mean provincial revascularization ratio was 1.12, with >2-fold variation among institutions. Revascularization ratios by institution are shown in Figure 2. There was no difference between the revascularization ratios at teaching centers (median 1.09; interquartile range: 1.04 to 1.47; n = 9) and nonteaching centers (median 1.10; interquartile range: 0.84 to 1.48; n = 10) (p = 0.74).
Predictors of revascularization
Factors associated with an initial treatment strategy of revascularization are shown in Table 2. Increasing age was associated with lower odds of revascularization (odds ratio [OR]: 0.99; 95% confidence interval [CI]: 0.98 to 1.00), whereas female sex conferred higher odds, with an OR of 1.16 (95% CI: 1.04 to 1.29). Medical comorbidities such as dialysis, previous stroke and peripheral vascular disease, and a history of CABG were all significant negative predictors of revascularization.
High-risk stress testing (electrocardiography and functional imaging) was not significantly associated with revascularization. However, declining LV function was strongly associated with an initial strategy of medical management only. Lesions in all coronary arteries were significantly associated with revascularization, but the ORs were highest for left main, proximal, and mid left anterior descending coronary artery disease. Although hospital characteristics were not associated with treatment strategy, angiography by an interventional cardiologist was significantly associated with higher odds of revascularization (OR: 1.46; 95% CI: 1.33 to 1.61).
Drivers of between-hospital variation
Reflecting the 2-fold variation in revascularization ratio, the MOR for the null model was 1.25 (Online Table 3). To put this value in context, in comparison with the majority of the ORs that predict revascularization in Table 2, this MOR was of greater magnitude. This suggests that the unexplained between hospital variation was as relevant as the majority of patient-, physician-, and hospital-level characteristics, with the exception of coronary anatomy. When patient, physician, and hospital level factors were incorporated, the MOR did not decrease, suggesting that the between hospital variation was not accounted for by the factors in our model.
Association with death and myocardial infarction
At last follow-up, 863 patients (18.6%) had died and 304 (6.5%) had experienced nonfatal myocardial infarctions in the medical management group, compared with 625 (11.9%) and 198 (3.8%), respectively, in the revascularization group (p < 0.001 for both). Outcomes over time are shown in Figure 3. An initial management strategy of revascularization was associated with a reduced hazard of myocardial infarction (hazard ratio [HR]: 0.58; 95% CI: 0.46 to 0.73) and death (HR: 0.81; 95% CI: 0.69 to 0.96), compared with medical therapy alone (Figure 4; full models shown in Online Table 4). This effect was consistent in all patients regardless of whether they underwent PCI or CABG (Online Table 5). Similarly, this effect was also consistent in the landmark analysis (Online Table 6).
When incorporating the interaction term of hospital revascularization tertile, these effects were also largely consistent across all tertiles of the revascularization ratio. High, medium, and low ratios were associated with significant reductions in the hazard of myocardial infarction (Figure 4). The effect on mortality was numerically consistent across tertiles but was statistically significant only at high-ratio centers. In addition, there was no evident relationship between hospital revascularization ratio and adjusted risk for myocardial infarction and death (Online Figure 1).
We found that the benefit of revascularization was consistent in patients with and those without LV dysfunction. The HR for death in patients with LV dysfunction (ejection fraction <50%) was 0.79 (95% CI: 0.65 to 0.95; p = 0.01), and the HR was 0.81 (95% CI: 0.64 to 1.02; p = 0.08) in those with preserved LV function. Similarly, when considering myocardial infarction, the HRs were 0.67 (95% CI: 0.51 to 0.87; p = 0.003) and 0.51 (95% CI: 0.38 to 0.68; p <0.001) for LV dysfunction and preserved LV function, respectively.
In this population-based study of patients undergoing coronary angiography for SIHD, we identified 2-fold institutional variation in revascularization of asymptomatic patients with obstructive coronary artery disease. As expected, the strongest predictor of revascularization was coronary anatomy; however, after accounting for patient, physician, and hospital factors, substantial variation in hospital revascularization practices remained. Despite this practice variation, we found an improvement in clinical outcomes associated with revascularization that was consistent across tertiles of hospitals.
Consistent with our hypothesis, we observed substantial variation in institutional practice in the treatment of asymptomatic patients. Previous work in a similar cohort consisting of patients with symptomatic SIHD demonstrated comparable levels of variation, but it was largely explained by patient-level factors (11). In contrast, the variation observed in our cohort was not due to the factors we modeled, despite including detailed patient characteristics, suggesting that institutional culture may be a significant contributor. Importantly, this variation per se did not translate into significant disparity in outcomes, as hospitals with high, medium, and low revascularization ratios had similar hazards of myocardial infarction and death. Although we hypothesized that careful patient selection at hospitals with low tertiles of revascularization would translate into an isolated improvement in outcomes with revascularization, no such relationship was observed. The consistency across hospital tertiles suggests that it is not an effect modifier, and higher ratios of revascularization at low-tertile hospitals may be appropriate.
The lack of an explanation for the practice variation we observed may have been driven by unmeasured confounders. For example, history of CABG was associated with a significantly lower likelihood of revascularization, suggesting that additional unmeasured anatomic factors may be important. This finding may be explained by the higher likelihood that these patients would have more diffuse disease and calcified vessels that would require complex intervention. Such interventions would be difficult to justify in the setting of an asymptomatic patient. Other factors, such as the presence of chronic total occlusions and measures of disease complexity (such as SYNTAX [Synergy Between PCI With Taxus and Cardiac Surgery] scores) may have been important in driving revascularization decisions.
The most important drivers of revascularization were related to patient characteristics, whereby higher levels of comorbid conditions and LV dysfunction were negative predictors. These findings suggest the presence of a “risk-treatment” paradox whereby higher risk patients are less likely to receive therapy that might alter their prognosis. However, clinical decision making in this population may be more difficult given the uncertain benefit of intervention and the absence of symptoms that drive management decisions. It is therefore reassuring that high-risk anatomic features such as left main or proximal left anterior descending coronary artery disease were strong predictors of revascularization, suggesting that clinicians are making decisions in line with disease distributions most likely to be associated with prognostic benefit.
Revascularization decisions are often driven by anginal symptoms, which were absent in our population, suggesting that the indication for intervention was the opportunity to alter prognosis. Unfortunately, little evidence is available to guide clinical decision making in asymptomatic patients, as only small subsets of these patients were included in previous studies. For example, the COURAGE trial included just 248 asymptomatic patients (3). Meanwhile, the best available evidence is from the SWISSI II (Swiss Interventional Study on Silent Ischemia Type II) trial, which randomized post–myocardial infarction patients with evidence of silent ischemia to PCI or medical therapy (16). Although that study showed a reduction in major adverse cardiac events associated with PCI, it was limited by a small population (n = 201) and noncontemporary treatment that did not involve stenting or statins. Retrospective observational studies have offered additional insights into the effect of PCI in SIHD and have generally shown improved outcomes with PCI (6–8). In our large population-based dataset, we found significant reductions in myocardial infarction and death associated with revascularization, and these findings were consistent with both PCI and CABG. These findings suggest that a selective revascularization strategy based on factors (such as high-risk anatomic distributions) most likely to confer prognostic benefit is safe and, combined with optimal medical therapy, may offer the best clinical outcome. However, these findings should be considered hypothesis generating and not be used as justification for routine revascularization of all coronary stenosis in asymptomatic patients. Further study is needed to help clinicians identify patients best suited to each management strategy.
First, our study was restricted to asymptomatic patients who nonetheless underwent coronary angiography, an atypical group for invasive investigation. We do not have information regarding the clinical decision that prompted angiography. As such, it is important that our findings not be extended to the larger group of asymptomatic patients with SIHD who do not undergo invasive investigations.
Second, our study’s observational design was susceptible to selection bias, as treatment assignment was not randomized, and we could not account for any potential unmeasured confounders. These confounders could include anatomic details (diffuse disease, calcification, and so on) as well as clinical factors such as patient preferences. The fact that we were not able to explain the variation in revascularization practices observed suggests that these may have played a role. As such, our results should be considered hypothesis generating. Looking forward, the ISCHEMIA (International Study of Comparative Health Effectiveness With Medical and Invasive Approaches) (NCT01471522) will likely offer additional insights in this patient population.
Third, our control group was not necessarily on optimal medical therapy. We did not have medication data on all patients but rather only on those older than 65 years. It is reassuring that although small differences in use were apparent between groups, generally high levels of medication use were achieved in all patients.
Fourth, although our overall population-based cohort was large, we assessed variation across a relatively small number of institutions, and as such our analyses regarding outcomes across hospital tertiles may have been underpowered.
Fifth, we also note that there was nearly 11% crossover from the medical group to revascularization, which biased our results to the null resulting in an underestimate of the potential treatment effect.
Finally, we do not have data regarding whether death was cardiac or noncardiac in nature, but given the subjective nature of this determination, the use of all-cause mortality in our study should reduce any potential bias related to the outcome.
Asymptomatic patients with SIHD and significant coronary obstruction continue to pose a therapeutic challenge to clinicians, as reflected by substantial practice variation. Although we found that the variation was difficult to explain, and it did not have clinical consequences, we found that a selective strategy of revascularization may be of benefit.
WHAT IS KNOWN? Revascularization in patients with SIHD is often based on the presence of symptoms, but little is known about the appropriateness of PCI or CABG in asymptomatic patients and the drivers of variation in treatment strategies.
WHAT IS NEW? We demonstrated 2-fold variation in revascularization practices that was not explained by measured patient-, physician-, or hospital-level factors. Despite this wide variation, a consistent reduction in myocardial infarction and death was observed in patients treated with revascularization.
WHAT IS NEXT? Further prospective and randomized studies are needed to determine which patients are most likely to benefit from selective revascularization and to reduce the variation observed in clinical practice.
The authors acknowledge that clinical registry data used in this publication are from participating hospitals through CorHealth Ontario, which serves as an advisory body to the Ontario Ministry of Health and Long-Term Care, is funded by the Ministry of Health and Long-Term Care, and is dedicated to improving the quality, efficiency, access, and equity in the delivery of the continuum of adult cardiac services in Ontario. The opinions, results, and conclusions reported in this paper are those of the authors and are independent from the funding sources. No endorsement by the Institute for Clinical Evaluative Sciences or the Ontario Ministry of Health and Long-Term Care is intended or should be inferred. Parts of this material are based on data and/or information compiled and provided by the Canadian Institute for Health Information. However, the analyses, conclusions, opinions, and statements expressed in the material are those of the authors and not necessarily those of the Canadian Institute for Health Information.
This work was supported by a grant-in-aid from the Heart and Stroke Foundation of Canada. Dr. Wijeysundera is supported by a Distinguished Clinician Scientist Award from the Heart and Stroke Foundation of Canada. Dr. Ko is supported by a Clinician Scientist Award from the Heart and Stroke Foundation of Canada, Ontario Provincial Office. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- coronary artery bypass graft
- confidence interval
- hazard ratio
- left ventricular
- median odds ratio
- odds ratio
- percutaneous coronary intervention
- stable ischemic heart disease
- Received May 30, 2018.
- Revision received September 19, 2018.
- Accepted October 30, 2018.
- 2019 American College of Cardiology Foundation
- ↵National Center for Health Statistics. Health, United States 2016: With Chartbook on Long-Term Trends in Health. Hyattsville, MD; 2017. Available at: https://www.cdc.gov/nchs/data/hus/hus16.pdf#019. Accessed November 8, 2018.
- Amsterdam E.A.,
- Wenger N.K.,
- Brindis R.G.,
- et al.
- Windecker S.,
- Stortecky S.,
- Stefanini G.G.,
- et al.
- Min J.K.,
- Berman D.S.,
- Dunning A.,
- et al.
- Hannan E.L.,
- Samadashvili Z.,
- Cozzens K.,
- et al.
- Patel M.R.,
- Calhoon J.H.,
- Dehmer G.J.,
- et al.
- Tu J.V.,
- Ko D.T.,
- Guo H.,
- et al.
- Bennell M.C.,
- Qiu F.,
- Kingsbury K.J.,
- Austin P.C.,
- Wijeysundera H.C.
- Oxner A.,
- Elbaz-Greener G.,
- Qui F.,
- et al.
- ↵Cardiac Care Network of Ontario. Annual Report 2015-16. Available at: https://www.corhealthontario.ca/CCN-Annual-Report-2015-16.pdf. Accessed November 8, 2018.
- Merlo J.,
- Chaix B.,
- Ohlsson H.,
- et al.