Long-Term Clinical Outcome in the Bypass Angioplasty Revascularization Investigation Registry
Comparison With the Randomized Trial
Background—The Bypass Angioplasty Revascularization Investigation (BARI) included 4039 patients with multivessel coronary artery disease; 1829 consented to randomization, and 2010 did not but were followed up in a registry. Thus, we can evaluate the outcome of physician-guided versus random assignment of percutaneous transluminal coronary angioplasty (PTCA) versus coronary artery bypass graft surgery (CABG).
Methods and Results—We compared the baseline features and outcomes for PTCA and CABG in the overall registry and its predesignated subgroups. We assessed the impact of treatment by choice versus random assignment by comparing the results in the registry with those of the randomized trial. Statistical adjustments for differences in baseline characteristics were made. Within the registry, nearly twice as many patients were selected for PTCA (1189) as CABG (625); mortality at 7 years was similar for PTCA (13.9%) and CABG (14.2%) (P=0.66) before and after adjustment for baseline differences between patients selected for PTCA versus CABG (adjusted RR, 1.02; P=0.86). In contrast to the randomized trial, the 7-year mortality rate of treated diabetics in the registry was equally high (26%) with PTCA or CABG. Seven-year mortality was higher for patients undergoing PTCA in the randomized trial than in the registry (19.1% versus 13.9%, P<0.01) but not for those undergoing CABG (15.6% versus 14.2%, P=0.57). The adjusted relative mortality risk for PTCA in the randomized versus registry population was 1.17 (P=0.16).
Conclusions—BARI physicians were able to select PTCA rather than CABG for 65% of registry patients who underwent revascularization without compromising long-term survival either in the overall population or in treated diabetics.
Randomized clinical trials are excellent vehicles for comparing therapeutic options, because randomization minimizes known and unknown biases and leads to balanced distribution of baseline variables between treatment groups.1 2 Although such trials provide valid comparison of therapies, the generalizability of results must be analyzed cautiously. Specifically, when a significant proportion of trial-eligible patients refuse randomization, and they differ from those actually randomized, the results of the trial may not be applicable to the entire population. Even when treatments result in equal outcome for the overall population, there may be differential effects in subsets of patients, such that physician judgment could result in a more optimal allocation of treatments. For many randomized trials, limited data collection on refusing patients precludes analysis of the former issue, whereas an inadequate sample size and lack of predesignation of subgroups precludes the latter. The Bypass Angioplasty Revascularization Investigation (BARI) was designed in scope and size to address both issues, because comprehensive baseline and follow-up data were collected not only on patients in the randomized trial but also in a registry of clinically and angiographically eligible patients who refused randomization.3
As we have previously reported, in the randomized trial, 5 years after enrollment, there was no significant difference in mortality between patients with multivessel coronary artery disease randomly assigned to an initial revascularization strategy of coronary artery bypass surgery (CABG) versus balloon angioplasty (PTCA).4 However, diabetic patients receiving oral hypoglycemic drugs or insulin at the time of study entry had significantly better survival when assigned to CABG than PTCA.4 5 Here, we report the outcome of PTCA versus CABG within the BARI Registry. By comparing the results with those of the randomized trial, we can assess the impact of physician-guided versus random allocation of therapy.
The BARI protocol, baseline characteristics of patients enrolled, and 5-year outcome data on the randomized population have been reported.3 6 7 BARI was a clinical trial designed to test the hypothesis that an initial strategy of PTCA rather than CABG did not compromise 5-year survival among patients angiographically suitable for either procedure who needed revascularization for multivessel coronary artery disease. Patients were enrolled at 18 sites throughout North America from August 1988 to August 1991. Patients who were clinically and angiographically eligible for the randomized trial but refused randomization while consenting to follow-up were entered into an “eligible not randomized” (ENR) registry. All patients received intensive medical therapy. Identical baseline data were obtained for both groups of patients. All patients had coronary arteriograms; the randomized patients had readings done at a central angiographic laboratory as well as at the clinical site, whereas the registry patients had site readings only. For comparability, this report uses site readings for both groups. Coronary anatomic and procedural characteristics and in-hospital outcomes were collected for all patients. Definitions of “significant” lesions and PTCA success have been provided previously.8 Both the randomized and the registry groups had follow-up data collected at 6 months, 1 year, and annually thereafter. Data in the randomized trial were obtained during alternate annual clinic visits or telephone interviews, whereas for the registry, data were obtained by telephone only. Vital status as of September 15, 1997, was collected for randomized and registry patients.
Treatment was defined according to the intention-to-treat principle for randomized patients. For registry patients, treatment was defined as the first revascularization procedure received during the initial 3 months after enrollment. In contrast to the randomized patients, the registry PTCA group includes a few patients (n=42) in whom other percutaneous technologies were used. Registry patients who had not been revascularized 3 months after enrollment were classified as the medical group.
Baseline characteristics of patients in the randomized trial and those in the registry were compared. Within the registry, 2 comparisons were made: patients who were revascularized versus those who were not, and patients who received PTCA versus CABG. χ2 tests, t tests, and Wilcoxon nonparametric tests were used for baseline comparisons. Procedural and in-hospital events were compared between patients in the randomized cohort and those in the registry within the PTCA and CABG groups. χ2 and t tests were used for procedural comparisons, and Fisher’s exact tests were used for comparing in-hospital complications. Long-term survival was compared for PTCA versus CABG within the registry and within the BARI a priori subgroups for the registry (defined by these baseline characteristics: number of diseased vessels, presence of class C lesions, classification of angina, and normal/abnormal left ventricular [LV] function) and the subgroup of treated diabetics, defined as patients receiving oral hypoglycemic drugs or insulin at enrollment. This subgroup was designated early in the study by the BARI Data Safety Monitoring Board for subgroup analysis and is consistently examined in BARI publications.4 7 Long-term survival comparisons for the randomized versus the registry patients were conducted for the entire study group and for these subgroups. As in the randomized trial, a value of P<0.05 was considered significant for comparisons of treatment or study type, and a value of P≤0.01 was considered significant for comparisons within a subgroup.
Seven-year mortality rates were computed from Kaplan-Meier survival curves.9 Unadjusted relative risks for mortality were calculated from univariate Cox proportional hazards regression models.10 Adjusted relative risks were calculated for PTCA versus CABG in the registry by multivariate Cox proportional hazards regression models, which control for factors significantly (P<0.05) associated with treatment selection (clinical site, age, race, history of smoking, hypertension, treated diabetes, self-rated health, duration of angina, number of diseased vessels, number of significant lesions, diffuse disease, class C lesions, and proximal left anterior descending coronary artery [LAD] disease). Similarly, the multivariate models for comparing the registry and randomized groups adjust for factors significantly associated with entry into the registry (clinical site, race, education, smoking history, self-rated health, activity level, history of congestive heart failure, duration of angina, proximal LAD disease, class C lesions, dominance, normal/abnormal ejection fraction, prior myocardial infarction, treated diabetes, and angina classification). In subgroup analyses, clinical site was not included as a covariate. All patients were included in the multivariate models regardless of missing covariate data. Missing values for the self-rated health variable (missing for 241 patients: 238 in the registry and 3 in the randomized cohort) were inferred from other baseline variables. Because missing LV function data (n=504 patients: 294 in the registry and 210 in the randomized trial) were associated with a higher mortality rate, an indicator variable denoting that this variable was missing was added to the multivariate model.11 Other missing covariate values were infrequent and were replaced with the mean value for that covariate. Rates of subsequent revascularization were compared between the randomized and registry populations by use of Kaplan-Meier curves with log-rank statistics and multivariate Cox proportional hazard models adjustment for the factors associated with entry in the registry.
Screening and Treatment Received
As shown in Figure 1⇓, 1829 of the 4107 patients eligible for BARI consented to random assignment to an initial strategy of CABG versus PTCA, 2010 refused randomization but consented to participation in the registry, and 268 refused both random assignment and follow-up in the registry.
In the registry, 80% of the patients were revascularized within 2 weeks and 90% within 3 months after enrollment. Within 3 months, 625 patients (31%) had received an initial CABG, 1189 (59%) had received an initial PTCA, and 192 (10%) had not been revascularized. In addition, 3 patients died and 1 withdrew within 3 months without having been revascularized. Thus, 196 registry patients who were not revascularized within 3 months constitute the medical group.
Within the randomized trial, 914 patients were assigned to CABG and 915 were assigned to PTCA. According to the protocol, these patients were to receive their assigned treatment within 2 weeks of randomization; this occurred in 89% of the patients. Three months after randomization, 98% of patients received their assigned revascularization procedure.
Baseline Characteristics for Study Groups
Baseline characteristics of randomized and registry patients are presented in Table 1⇓. Numerous differences existed between the 2 study groups. The randomized population included more minorities; more patients with a high school education or less, lower activity levels, and poorer self-rated health status; and more current smokers. The randomized patients had a greater proportion of patients with a history of congestive heart failure (9% versus 5%, P<0.001), treated diabetes (19% versus 17%, P=0.05), angina for >1 year (44% versus 37%, P<0.001), significant proximal LAD lesions (40% versus 35%, P=0.004), and abnormal ejection fractions (20% versus 16%, P=0.002). However, the randomized and registry groups were similar in terms of age, sex, proportion of patients with triple-vessel and peripheral vascular disease, and average number of significant lesions. The percentage of eligible patients enrolled in the randomized trial as opposed to the registry varied greatly among the clinical sites, ranging from 28% to 94%.
Table 2⇓ shows the association between baseline factors and the selection of PTCA, CABG, or medical therapy within the registry. Ninety percent of registry patients (n=1814) underwent revascularization. These patients had a higher prevalence of unstable angina and significant proximal LAD disease than those who were not revascularized (n=196). Age, sex, and race were not significantly associated with undergoing revascularization.
Among the 90% of registry patients who were revascularized, the proportion that was selected for CABG (34%) versus PTCA (66%) varied substantially among the clinical sites, ranging from 12% to 75%. CABG patients were slightly older (62.5 versus 61 years, P<0.001), more frequently had angina for >1 year (42% versus 34%, P<0.001), and less frequently had a smoking history (64% versus 70%, P<0.001). CABG patients had more severe coronary angiographic profiles than PTCA patients, with more significant lesions (3.6 versus 3.0, P<0.001), and a higher prevalence of triple-vessel disease (51% versus 30%, P<0.001), significant proximal LAD lesions (46% versus 32%, P<0.001), any class C lesion (44% versus 35%, P<0.001), and diffuse disease (36% versus 26%, P<0.001). An abnormal LV ejection fraction was infrequent and similar in the CABG and PTCA registry patients.
Initial Revascularization Procedure (Registry Versus Randomized Trial)
Table 3⇓ compares patients having PTCA or CABG according to whether they were in the randomized trial or the registry. Patients undergoing CABG in the randomized trial had fewer significant lesions (3.2 versus 3.6, P<0.001), received fewer grafts (2.8 versus 3.1, P<0.001), and had a shorter bypass time (91 versus 100 minutes, P<0.001) and a lower likelihood of use of the internal mammary artery as a conduit (82% versus 88%, P=0.002) than registry CABG patients. The high percentage of patients with all intended vessels grafted and leaving the operating room in stable condition, and low rates of Q-wave myocardial infarction and mortality in-hospital did not differ significantly between randomized and registry CABG patients.
Patients undergoing PTCA in the randomized trial had more significant lesions (3.2 versus 3.0, P<0.001) and lesions attempted (2.3 versus 1.9, P<0.001) than registry PTCA patients; however, the percentage of lesions attempted did not differ significantly. The percentage of patients with all attempted lesions successfully dilated (73% versus 78%, P=0.019) was lower in randomized than registry patients. A slightly lower percentage of randomized than registry patients (95% versus 97%, P=0.024) left the catheterization laboratory in stable condition, and more randomized than registry patients required emergency CABG (6.3% versus 4.1%, P=0.027). Neither the occurrence of Q-wave myocardial infarction nor in-hospital mortality, which was ≈1%, differed significantly between PTCA patients in the randomized trial or registry.
Survival for registry patients is displayed in Figure 2⇓ according to initial revascularization strategy: PTCA, CABG, or medical therapy. Seven-year mortality rates were 14.2% for CABG, 13.9% for PTCA, and 19.2% for the medical group. There was no significant difference in unadjusted mortality among the 3 groups (P=0.09) or between the PTCA and CABG groups (P=0.66). According to a multivariate model, the adjusted relative risk for PTCA versus CABG was 1.02; 95% CI, 0.79 to 1.33; P=0.86. As shown in Figure 3⇓, 7⇓-year mortality in the registry also did not differ significantly between PTCA and CABG in any of the subgroups prespecified by the BARI protocol. However, the adjusted RR for PTCA versus CABG in the subset of patients with stable angina was 2.36 (99% CI, 1.12 to 4.97; P=0.0029).
Survival curves for patients in the randomized trial compared with those in the registry are shown in Figure 4⇓. The 7-year mortality rate was 17.3% in the randomized cohort versus 14.5% in the registry (P=0.14). After adjustment for baseline factors that were significantly different between registry and randomized patients, the relative risk was 1.02 (P=0.85) for the randomized trial versus registry.
Figure 5⇓ shows survival for randomized versus registry patients by initial revascularization strategy (CABG or PTCA). Seven-year mortality rates for randomized patients assigned to CABG did not differ significantly from those selected for CABG in the registry (15.6% versus 14.2%; unadjusted RR, 1.08; P=0.57; adjusted RR, 0.94; P=0.66). In contrast, randomized patients assigned to PTCA had higher mortality rates than registry patients selected for PTCA (19.1% versus 13.9%; unadjusted RR, 1.43; P<0.01); this difference was not statistically significant after adjustment (adjusted RR, 1.17; P=0.16).
When the subgroups predesignated by the BARI protocol were examined individually, the randomized and registry CABG patients had similar survival in each subgroup before and after adjustment for baseline differences between randomized and registry patients. For PTCA, randomized patients had significantly poorer survival than registry patients for the subgroups with unstable angina, abnormal LV function, and no class C lesions. After adjustment for differences in baseline variables between the randomized and registry populations, only the subgroups of randomized PTCA patients with treated diabetes (adjusted RR, 1.54; P=0.023) and unstable angina (adjusted RR, 1.32; P=0.033) approached significance compared with registry patients.
Subsequent Revascularization Procedures
Among patients who underwent an initial CABG, the needs within 7 years for any subsequent revascularization procedure (13.1% versus 12.0%, P=0.48) or subsequent CABG in particular (1.7% versus 2.3%, P=0.45) were similar between randomized and registry patients. As shown in Figure 6⇓, among patients who underwent an initial PTCA, the randomized and registry patients had a similar rate of any subsequent revascularization procedure within 7 years (59.7% versus 58.7%, P=0.43) and subsequent CABG in particular (35.5% versus 34.0%, P=0.52), respectively. These differences remained insignificant after adjustment for differences between the randomized and registry populations.
As shown in Figure 7⇓, angina pectoris, which was present in >90% of patients at baseline, was infrequent at each follow-up interval for randomized or registry patients in either revascularization strategy. Registry PTCA patients were significantly less likely to have angina than randomized PTCA patients at 1, 3, and 5 years after enrollment (15.8% versus 20.3% at 5 years, P=0.01).
Treatment Selection and Long-Term Outcome in the Registry
The BARI Registry allows a unique overview of the selection process and long-term outcome of PTCA versus CABG in patients with PTCA-eligible multivessel coronary artery disease in the era preceding the widespread use of stents and glycoprotein IIb/IIIa receptor blockers. Within the registry, PTCA was clearly the initial revascularization strategy favored by patients and their physicians, being chosen nearly twice as often as CABG. Reflecting clinical practice at that time, CABG was generally reserved for patients with more lesions, proximal LAD disease, and type C lesions, as had been observed in the EAST registry.12
Within the BARI registry, there was no significant difference in 7-year mortality between PTCA and CABG before or after adjustment for baseline differences in either the overall population or any of the subgroups analyzed, with the exception of patients with stable angina. This finding of a higher risk for PTCA in this subgroup had not been observed in previous reports or in our randomized cohort and may have resulted from the play of chance. The subgroup of registry patients with treated diabetes had similar 7-year survivals with PTCA or CABG, both before and after statistical adjustment. A detailed analysis of this subgroup is the subject of a separate publication.13
Comparison of Long-Term Outcome in the Registry and Randomized Populations
Registry patients had better crude long-term survival than those in the randomized trial, particularly when PTCA was the initial revascularization strategy. Although the clinical and angiographic criteria for inclusion in the registry and randomized trial were identical, resulting in similar patient profiles, there were some significant differences between the registry and randomized populations. After adjustment for these baseline clinical and angiographic differences, the overall survival differences were no longer significant.
Among certain subgroups, such as treated diabetics and patients with unstable angina, abnormal LV function, and no type C lesions, those undergoing PTCA had significantly better survival in the registry than the randomized trial. Even after statistical adjustments, strong trends toward greater survival in registry versus randomized PTCA patients remained for treated diabetics and patients with unstable angina.
In analyzing how interventional cardiologists were able to obtain better long-term survival in the registry population than the randomized trial, while selecting PTCA for 65% of the patients who were revascularized, we can evaluate objective criteria by which registry patients were selected for PTCA. By selecting patients with fewer significant and type C lesions for PTCA in the registry versus randomized population, interventional cardiologists in BARI were able to achieve slightly but significantly higher success rates, in terms of the percentage of patients in whom all lesions attempted were successfully dilated.14 This may have also resulted in the lower rates of angina in registry versus randomized PTCA patients at 1, 3, and 5 years.
By selecting angiographically favorable patients for PTCA, survival and the need for subsequent revascularization procedures might have been adversely affected in CABG patients within the registry. However, important predictors of decreased survival, such as advanced age and decreased ejection fraction, did not predominate in patients selected for CABG.15 16 The factors that did lead to a choice of CABG in the registry, such as a greater number of significant lesions, type C lesions, and proximal LAD disease, are not powerful predictors of an adverse outcome with surgery.16 Instead, they would predict more grafts, a greater likelihood of an internal mammary artery graft, and a slightly longer time on cardiopulmonary bypass, with no significant difference in mortality or morbidity, which is what was observed.
Although many clinical and angiographic characteristics were measured in BARI, it is possible that others were not or could not be measured. Because the registry is a self-selected group, it is possible that the patient who refused randomization was more likely to practice better health-related behavior. Therefore, although we controlled for measured differences between the 2 study populations, we realize that this statistical adjustment is imperfect. Another limitation of this study is that 10% of registry patients were not revascularized. Hence, the revascularized patients represent a biased subset of the registry, and it is difficult to draw meaningful conclusions regarding the medical group. Finally, intracoronary stenting and the administration of glycoprotein IIb/IIIa inhibitors, which reduce restenosis and procedural complications, respectively, were not evaluated in BARI.17 18
The BARI registry and randomized groups represent the largest population of PTCA-eligible patients with multivessel coronary artery disease with prospectively gathered comprehensive clinical and angiographic data in whom the selection of revascularization of CABG versus PTCA by choice versus random assignment can be evaluated. This analysis indicates that using clinical and angiographic criteria, we were able to select a significant majority of the registry population for initial revascularization by PTCA without compromising long-term survival in either the overall population or patients treated for diabetes.
This study was supported by grants HL-38493, HL-38504, HL-38509, HL-38512, HL-38514-6, HL-38518, HL-38524-5, HL-38529, HL-38532, HL-38556, HL-38610, HL-38642, and HL-42145 from the National Heart, Lung, and Blood Institute, Bethesda, Md.
- Received April 4, 1999.
- Revision received December 12, 1999.
- Accepted January 25, 2000.
- Copyright © 2000 by American Heart Association
Friedman LM, Furberg C, DeMets DL. Fundamentals of Clinical Trials. 2nd ed. New York, NY: Raven Press; 1985:51.
Pocock SJ. Clinical Trials: A Practical Approach. New York, NY: John Wiley & Sons; 1983:50.
The BARI Investigators. Protocol for the Bypass Angioplasty Revascularization Investigation. Circulation. 1991;84(suppl V):V-1–V-27.
The Bypass Angioplasty Investigation (BARI) Investigators. Influence of diabetes on 5-year mortality and morbidity in a randomized trial comparing CABG and PTCA in patients with multivessel disease. Circulation. 1997;96:1761–1769.
The Writing Group for the Bypass Angioplasty Revascularization Investigation (BARI) Investigators. Five-year clinical and functional outcome comparing bypass surgery and angioplasty in patients with multivessel coronary disease: a multicenter randomized trial. JAMA. 1997;277:715–721.
Cox DR. Regression models and life-tables. J R Stat Soc (B). 1972;34:187–220.
Rubin D. Influence of missing data. Biometrika. 1976;63:581–592.
Detre KM, Guo P, Holubkov R, et al. Coronary revascularization in diabetic patients: a comparison of the randomized and observational components of the Bypass Angioplasty Revascularization Investigation (BARI). Circulation. 1999;99:633–640.
Ryan TJ, Bauman WB, Kennedy JW, et al. ACC/AHA Task Force Report. Guidelines for percutaneous transluminal coronary angioplasty: a report of the American College of Cardiology/American Heart Association Task Force on Assessment of Diagnostic and Therapeutic Cardiovascular Procedures (Committee on Percutaneous Transluminal Coronary Angioplasty). J Am Coll Cardiol. 1993;22:2033–2054.
Gersh BJ, Kronmal RA, Frye RL, et al. Coronary arteriography and coronary artery bypass surgery: morbidity and mortality in patients 65 years or older: a report from the Coronary Artery Surgery Study. Circulation. 1983;67:483–491.