Three-Year Outcome After Coronary Stenting Versus Bypass Surgery for the Treatment of Multivessel Disease
Background— The primary results of Arterial Revascularization Therapy Study reported a greater need for repeated revascularization after percutaneous coronary intervention with stenting (PCI). However, PCI was less expensive than coronary artery bypass grafting (CABG) and offered the same degree of protection against death, stroke, and myocardial infarction.
Methods and Results— Patients with multivessel disease (n=1205) were randomly assigned to either CABG or PCI and followed up for up to 3 years. Survival rates without stroke or myocardial infarction were similar in each group at 1 year and 3 years (90.5% versus 91.4% for PCI versus CABG at 1 year and 87.2% versus 88.4% for PCI versus CABG at 3 years). However, the respective repeat revascularization rates were 21.2% and 26.7% at 1 and 3 years in patients allocated to PCI, compared with 3.8% and 6.6% in patients allocated to CABG (P<0.0001). Diabetes (P<0.0009) and maximal pressure for stent deployment (P<0.002) are the strongest independent predictors of events at 3 years after PCI, whereas left anterior descending coronary artery grafting (P<0.006) is the best predictor of event-free survival at 3 years after CABG. The incremental cost of surgery compared with PCI for an event-free patient was 19 257 at 1 year but decreased to 10 492 at 3 years. It remained at 142 391 at 3 years when revascularization procedures were excluded in the efficacy end point, however.
Conclusions— Three-year survival rates without stroke and myocardial infarction are identical in both groups, and the cost/benefit ratio of stenting is determined primarily by the increasing need for revascularization in the PCI group.
Received February 4, 2003; de novo received October 7, 2003; revision received December 1, 2003; accepted December 2, 2003.
The Arterial Revascularization Therapy Study (ARTS) was designed to compare coronary artery bypass grafting (CABG) and stenting for the treatment of patients with multivessel coronary disease (MVD).1 One year after the procedure, coronary stenting demonstrated a substantial reduction in costs at the expense of a 14.4% difference in major adverse cardiac and cerebral events (MACCEs).
See p 1079
However, beyond the first year, the initial favorable cost-effectiveness of the percutaneous approach could decrease with time because the need for late repeat revascularization may be substantially higher after percutaneous procedures in patients with MVD. Indeed, a more rapid progression of the disease, particularly in some subsets of patients, such as diabetics, may have an unfavorable impact on the cohort of patients treated with percutaneous techniques compared with CABG.2 Moreover, a late protective effect of surgical revascularization is traditionally attributed to the more complete revascularization or to the fact that proximal diseased segments of epicardial vessels are permanently bypassed.3,4
This report presents the clinical outcome, quality of life, and cost-effectiveness at 3 years of the patients enrolled in the ARTS trial.
Between April 1997 and June 1998, 1205 patients were randomized to either stent implantation (n=600) or CABG (n=605).1 The study population includes 208 diabetic patients. The indications for revascularization included silent ischemia, stable or unstable angina pectoris, and the presence of at least 2 de novo lesions located in different major epicardial coronary arteries potentially amenable to stent implantation.
Specific exclusions from the randomized trial may be summarized as follows: left ventricular ejection fraction <30%, left main stenosis, history of cerebrovascular accident (CVA), transmural myocardial infarction (MI) within the preceding week, severe hepatic or renal disease, and need for concomitant major surgery.1,5 All patients gave written, informed consent.
Three-Year Clinical Follow-Up
The study protocol required all patients to have follow-up clinic visits with an ECG at 1, 2, and 3 years. At each visit, physical examination, anginal status, and use of medications were assessed. Additional information was obtained by telephone interview or via the referring physician when needed. An independent committee adjudicated clinical events and ECGs. Quality of life (EQ-5D) was assessed at 1 and 3 years.
Clinical End Points and Effectiveness
The primary end point was defined as the absence of any of the following MACCEs within 12 months after randomization: death, CVA, documented nonfatal MI adjudicated by either new abnormal Q wave or predefined enzymatic changes, or repeat revascularization by coronary stenting or CABG.1,5
Secondary objectives of the study were to compare both strategies at 3 years. MACCEs were counted from the time of randomization, whereas the clinical status and medications were assessed at predetermined times of 1, 2, and 3 years after the procedure. Of 1205 patients eligible for the trial, 1 patient was lost to follow-up, 3 were alive but withdrew their consent from further participation in the trial, and 2 patients were never treated by either modality.
Secondary measures of efficacy were assessed by means of the EQ-5D questionnaire, which allows patients to grade their general health status.6
Costs and Cost-Effectiveness
The costs that were considered were the direct medical costs per patient, calculated as the number of resource units used multiplied by the cost per unit. The resources included diagnostic tests, devices and material used, diagnostic or therapeutic procedures, hospital days, medications, and rehabilitation services. All physician visits and hospitalizations occurring during the 3-year follow-up were noted in the Case Report Form. Unit costs were based on estimates provided by the Dijkzigt Hospital, Rotterdam, the Netherlands, as reported previously.1,7
Statistical analysis was performed with SAS 6.12 software (SAS Institute Inc). Continuous variables were expressed as mean±SD and were compared by unpaired Student’s t test or Wilcoxon ranked scores when applicable. Fisher’s exact test was used for categorical variables. Binary outcome variables are reported as frequencies and percentages and were compared in terms of relative risk with 95% CIs calculated by the formula of Greenland and Robins.8 All analyses were based on the intention-to-treat principle, and statistical tests were 2 tailed.
Event-free survival was estimated by the Kaplan-Meier method, and differences were assessed by the log-rank test.
A 2-step approach was followed to identify predictors of the primary end point at 1 year and 3 years. First, univariate logistic regression was applied, including all clinical and angiographic baseline characteristics and procedure-related variables (including method of revascularization: ie, stenting or surgery). Variables that proved to be significant univariate predictors of MACCEs were entered into the multivariate logistic regression analysis. Within this, the final model was constructed on the basis of maximum-likelihood estimates using a stepwise selection method with an entry limit of 0.05. The predictive accuracy of the multivariate models is presented using the concordance index (c index).9 With regard to internal and external validation of the final model, no other steps were taken.
Table 1 displays the baseline and procedural characteristics of ARTS randomized patients. The randomized groups were similar with regard to demographic and anatomic characteristics. An equivalent anatomic degree of revascularization was achieved in each group.
Three-Year Clinical Outcome
At 1 year, patients assigned to stenting had an increased risk of repeat revascularization (21.0% versus 3.8%, P<0.001). Between 1 and 3 years, repeat revascularization with percutaneous or surgical techniques was performed more often after stenting than after CABG (9.1% versus 3.6%, P=0.0001), whereas a similar number of patients in the 2 groups died or suffered a stroke or an MI (Table 2, Figure 1).
After 3 years, patients in the surgery group had significantly less angina (12.8% versus 18.4% in the stenting group, P=0.011). The better symptomatic outcome after surgery was also reflected in a lower rate of use of antianginal medications (β-blockers, calcium antagonists, and/or nitrates used in 78.4% versus 65.4% at 3 years, P<0.001 among patients revascularized by stenting versus CABG).
Predictors of Late Clinical Outcome
We examined independent risk factors for MACCEs at 1 and 3 years for the patients allocated to stenting or CABG (Table 3). After stenting, the most significant positive predictor was high-pressure stent deployment and the most significant negative predictor was the presence of diabetic mellitus, both at 1 and 3 years. The other negative predictors of 3-year events included stenosis in the mid circumflex or in the distal dominant right coronary artery (RCA) or the use of digitalis. Previous MI conferred a protective effect at 3 years. After surgery, the most important predictor of a favorable outcome was an anastomosis in the mid left anterior descending artery (LAD). Conversely, intention to treat and incomplete revascularization of the distal RCA were associated with worse outcomes. It is worth noting that none of the variables that predicted 1-year outcomes in the surgical group were predictors of the outcomes at 3 years.
A total of 208 patients (17.3%) were diabetic (3.2% insulin dependent). At 3 years, the incidence of death, stroke, and MI was similar between patients with or without diabetes assigned to stenting or CABG (Table 4). However, in diabetics assigned to stenting, the need for repeat revascularization was higher within the first year and between 1 and 3 years (15.2% additional reinterventions in diabetics versus 7.7% in nondiabetics, P=0.023). As a consequence, the clinical outcome of diabetic patients treated with stenting was worse at 3 years (Figure 2).
Quality of Life
There were no differences in quality of life as assessed by the self-rated EQ-5D questionnaire between 1 and 3 years among patients allocated to stenting or bypass surgery. More specifically, the benefit observed after CABG in specific domains such as “mobility” and “anxiety or depression” at 1 year disappeared by 3 years (Table 5).
At 3 years, the additional costs generated by the higher rate of revascularization in the stented patients reduced the cost saving of 2779 in favor of stenting observed at 1 year to 1798 (Table 6).
Conceptually, a policy of performing elective CABG rather than angioplasty and stenting in every patient enrolled in the trial would have generated 1 additional event-free patient for 5.8 patients treated. Because the difference in costs between surgery and stenting is 1798 at 3 years, the incremental cost for each additional event-free patient treated by surgery amounts to 10 492 . A similar cost-effectiveness evaluation at 1 year yielded an incremental cost of 19 257 for each additional event-free patient treated by surgery. When the revascularization procedures were excluded within the efficacy end point, the additional costs per additional survivor free of MI and/or CVA are estimated at 307 145 at 1 year and 142 391 at 3 years (Table 6).
The results of ARTS at 3 years report identical survival without stroke and MI in both groups. However, the need for repeat revascularization is increasing in the percutaneous coronary intervention (PCI) group, notably in diabetic patients. The present study also provides greater clarity about the cost and clinical efficacy of coronary angioplasty and stenting as an alternative to CABG.
CABG Versus PCI
The meta-analysis of trials comparing an initial strategy of CABG or balloon angioplasty showed a much higher risk of repeat revascularization (33.7%) at 1 year after balloon angioplasty than the 1-year revascularization rate of 21.0% observed in ARTS.1,11–17 In addition, there was a trend toward an increased mortality rate at 3 to 5 years in the balloon angioplasty group that is not apparent in this study. Conversely, the 1-year rate of revascularization after surgery, which is 3.8% in ARTS, is similar to the 3.3% rate observed in these earlier trials. Beyond 1 year, rates of reintervention were 4.5 and 1.8 per 100 patient-years of follow-up with balloon angioplasty and CABG, respectively, a figure equal to the 4.5 and 1.8 per 100 patient-years in the ARTS stenting and surgery groups, respectively.16–18 Thus, the initial strategy of stenting is entirely defensible, because in most patients, it was a definitive treatment and did not merely delay bypass surgery.
Predictors of MACCEs
It is worth noting that in the stent group, the major predictors of MACCEs at 1 and 3 years (diabetes mellitus, disease involving the RCA) are similar, whereas higher balloon inflation pressure is associated with better outcomes at these time points. Conversely, in the surgical group, the major predictors of adverse outcomes at 1 year (elevated creatine kinase-MB, increasing age, use of heparin) are no longer predictive at 3 years. Instead, anastomosis to the LAD has a markedly positive impact on outcome, whereas intended or incomplete revascularization of the RCA has a negative prognostic significance.
Influence of Diabetes
In diabetic patients, the benefit of surgery over stented angioplasty is even more pronounced at 3 years than at 1 year.19 A mortality rate of 7.1% at 3 years was observed with stenting, versus 4.2% with CABG (P=0.39). However, this trend in favor of surgery needs to be assessed by a longer follow-up or larger sample size.
It is worth noting that the incidence of mortality from 1 to 3 years was similar in both groups (0.8% with stenting versus 1.1% with CABG, P=NS). In contrast, repeat revascularizations were almost twice as frequent both during the first year and beyond 1 year. The increased need for revascularization in diabetics during the first year may be a result of the more aggressive restenotic process, resulting in a greater intrinsic risk of in-stent restenosis.20 Beyond the first year, progression of underlying atherosclerosis may play a greater role. Whereas stenting treats a discrete lesion at a particular point in time, CABG provides an alternative conduit to the vessel so that any new lesion occurring proximal to the site of bypass may be inconsequential.
Quality of Life
The Randomised Intervention Treatment of Angina (RITA) and Coronary Angioplasty versus Bypass Revascularization Investigation (CABRI) trials showed no difference in health-related quality of life at 1 year after CABG or angioplasty.21,22 In the Bypass Angioplasty Revascularization Investigation (BARI) trial, there was a better functional status at 1 year among the patients undergoing CABG, but this difference had diminished after 4 years.23 We made a similar observation in ARTS: at 3 years, no significant differences exist between the 2 treatments, but with the exception of the pain and discomfort domain, all indexes were still more favorable in the CABG group. It must be recognized, however, that EQ-5D was developed to estimate usefulness and may not be as sensitive as disease-specific instruments to analyze treatment results. Nevertheless, EQ-5D captured differences between the results of treatment before and after the 2 procedures.1
Cost and Cost-Effectiveness
In trials seeking to compare cost-effectiveness of PCI, repeat revascularization is usually considered an adverse clinical event, although it is also an economic end point. Moreover, although it may be associated with short-term impairment in quality of life, its long-term prognostic importance is uncertain. In ARTS, as in other studies,24–27 the cost-effectiveness analysis includes repeat revascularization within the efficacy end point. Consequently, the incremental cost combined with more clinical events in patients assigned to stenting significantly reduced the relative cost-effectiveness of stenting after 3 years. Therefore, the incremental cost for each event-free patient treated by surgery decreased substantially from 19 257 at 1 year to 10 492 at 3 years. Here, it may be noted that the calculations are based on the costs that applied at the time of the study. Specifically, a decrease in the price of the stent, such as occurred during the past few years, may tip the balance more favorably toward stenting.
In addition, when revascularization procedures are excluded as an efficacy end point, the incremental cost of CABG remains substantial at 3 years and does not seem favorable for surgery unless there is a long-term survival benefit.
The 3-year results of the ARTS trial have identified predictors of untoward outcome after stenting or CABG for MVD. Bypass surgery seemed to be the preferred mode of therapy for patients with diabetes or distal RCA or circumflex stenoses, and high-pressure stent deployment should generally be recommended. After surgery, the clinical outcome is also better when a conduit is placed on the LAD but is worse if the RCA is not properly revascularized. Finally, although stenting remains less expensive at 3 years, surgery offers the most effective method of revascularization, with similar quality of life. Whether off-pump coronary surgery for stenosis of the LAD combined with stents to the other vessel or the use of drug-eluting stents could yield better results in the ARTS population is speculative. These new treatment modalities need to be addressed in the near future to define the best treatment strategy in MVD.
We thank Dr Brian G. Firth for his careful review of the manuscript and for his constructive suggestions.
Kendal DG. Rank Correlation Methods. 3rd ed. London, UK: Charles Griffin; 1962.
Cox DR. Fieller’s theorem and a generalization. Biometrika. 1967; 54: 567–572.
Rodriguez A, Boullon F, Perez-Balino N, et al. Argentine randomized trial of percutaneous transluminal coronary angioplasty versus coronary artery bypass surgery in multivessel disease (ERACI): in-hospital results and 1-year follow-up. ERACI Group. J Am Coll Cardiol. 1993; 22: 1060–1067.
Abizaid A, Costa MA, Centemero M, et al. Clinical and economic impact of diabetes mellitus on percutaneous and surgical treatment of multivessel coronary disease patients: insights from the Arterial Revascularization Therapy Study (ARTS) Trial. Circulation. 2001; 104: 533–538.
Kornowski R, Mintz GS, Kent KM, et al. Increased restenosis in diabetes mellitus after coronary interventions is due to exaggerated intimal hyperplasia: a serial intravascular ultrasound study. Circulation. 1997; 95: 1366–1369.
Pocock SJ, Henderson RA, Seed P, et al, for the RITA Trial Participants. Quality of life, employment status, and anginal symptoms after coronary angioplasty or bypass surgery: 3-year follow-up in the Randomised Intervention Treatment of Angina (RITA) trial. Circulation. 1996; 94: 135–142.
Währborg P, on behalf of the CABRI Trialists. Quality of life after coronary angioplasty or bypass surgery: 1-year follow-up in the Coronary Angioplasty versus Bypass Revascularization Investigation (CABRI) trial. Eur Heart J. 1999; 20: 653–658.
Serruys PW, de Bruyne B, Carlier S, et al. Randomized comparison of primary stenting and provisional balloon angioplasty guided by flow velocity measurement. Doppler End Point Balloon Angioplasty Trial Europe (DEBATE) II Study group. Circulation. 2000; 102: 2930–2937.
Suryapranata H, van’t Hof AW, Hoorntje JC, et al. Randomized comparison of coronary stenting with balloon angioplasty in selected patients with acute myocardial infarction. Circulation. 1998; 97: 2502–2505.
Cohen DJ, Taira DA, Berezin R, et al. Cost-effectiveness of coronary stenting in acute myocardial infarction: results from the Stent Primary Angioplasty in Myocardial Infarction (stent-PAMI) trial. Circulation. 2001; 104: 3039–3045.
Stone GW, Grines CL, Rothbaum D, et al. Analysis of the relative costs and effectiveness of primary angioplasty versus tissue-type plasminogen activator: the Primary Angioplasty in Myocardial Infarction (PAMI) trial. The PAMI Trial Investigators. J Am Coll Cardiol. 1997; 29: 901–907.