Similarity of Ventricular Function in Patients Alive 5 Years After Randomization to Surgery or Angioplasty in the BARI Trial
Background—Left ventricular ejection fraction (LVEF) is a recognized determinant of survival in patients with coronary artery disease. In major trials comparing surgical and percutaneous revascularization approaches, the long-term effect of the coronary revascularization strategy on LVEF has not been reported.
Methods and Results—In the NHLBI-sponsored Bypass and Angioplasty Revascularization Investigation (BARI) randomized trial comparing angioplasty and bypass surgery as initial treatment strategies, 1220 (75%) of the 1617 surviving randomized patients had their EF measured by radionuclide ventriculography 5 years after study entry. For the total study group, the 5-year EF in the CABG group (n=623) was 55.8±12.3, compared with 55.7±12.7 in PTCA group (n=597, P=0.82). There was no significant difference in measured EF between the CABG group and the PTCA group within multiple subgroups determined by the presence or absence of diabetes, 3-vessel disease, complete revascularization, or prior myocardial infarction. In a multiple linear regression model developed to predict 5-year EF, treatment assignment to PTCA or CABG was not significant (P=0.95). If an EF of 0 was imputed for patients who were dead and missing EF data, however, there was a higher EF in the CABG group (P=0.0018) among diabetic patients only.
Conclusions—In the BARI randomized trial, initial treatment assignment to angioplasty was not associated with any difference in long-term ventricular function compared with initial treatment assignment to surgery. These results apply, however, only to patients who were alive at 5 years.
Several trials with either coronary artery bypass graft surgery (CABG) or percutaneous transluminal coronary angioplasty (PTCA) have demonstrated comparable survival at 1 year,1 3 years,2 and 5 years3 after coronary revascularization.
Left ventricular ejection fraction (LVEF) is a recognized determinant of survival in patients with coronary heart disease.4 LVEF is a covariable of benefit derived by CABG.5 Although factors that may affect LV function after revascularization, including perioperative myocardial infarction (MI),2 3 cumulative MI,1 2 3 and subsequent repeat intervention,1 2 have been recorded in the major trials comparing surgical and percutaneous revascularization procedures, the long-term effect of coronary revascularization strategy (based on the intention to treat) on LVEF has not been described.
The NHLBI-sponsored BARI Multicenter Trial was designed primarily to test the hypothesis that long-term clinical outcome of patients with multivessel coronary disease suitable for treatment with either PTCA or CABG is not compromised when PTCA is chosen as the initial treatment strategy.3 6 Death and MI at 7 years after randomization were greater after initial treatment with PTCA than with CABG, a difference entirely attributable to diabetic patients on drug therapy.6 Baseline LV function was not associated with an adverse 5-year cardiac mortality or with the composite end point of cardiac mortality or MI.7
A large subset of the BARI population underwent serial LV function assessments with radionuclide ventriculography (RVG). The primary hypothesis of the present analysis is that initial treatment assignment to PTCA, compared with initial treatment assignment to CABG, is not associated with any difference in measured LVEF at 5 years, independent of the completeness of coronary revascularization, and the baseline presence or absence of diabetes, 3-vessel disease, and abnormal LV function. The present study reports the relationship among primary and secondary outcomes, baseline and 5-year RVG LV function, comorbidities (eg, diabetes), and coexistent coronary artery anatomy (before and after revascularization) in a large BARI patient subset followed up for 5 years after randomization to either CABG or PTCA.
BARI Trial and Study Group
A detailed description of the BARI study design, protocol, and clinical characteristics of the patients at entry has been published previously.3 4 8 Briefly, patients who had multivessel coronary artery disease, severe angina, or ischemia and were suitable for both PTCA and CABG were randomly assigned to an initial treatment strategy of CABG or PTCA. The primary clinical indications for coronary revascularization included (1) non–Q-wave MI or unstable angina stabilized for 4 hours to 6 weeks; (2) stable class III or IV angina pectoris; (3) stable class I or II angina with either severe ischemia on noninvasive testing, Q-wave MI stabilized for 24 hours to 30 days, or resting EF <50%; or (4) no angina during the 6 weeks before entry but objective evidence of severe ischemia on noninvasive testing and either prior Q-wave MI or history of prior angina. Patients were considered ineligible if any of the following criteria were present: single-vessel disease, age <17 or ≥80 years, prior coronary revascularization procedure, left main coronary stenosis ≥50%, or noncardiac illness expected to limit survival. From August 1988 through August 1991, 1829 patients consented to the randomization process; 915 were assigned to PTCA and 914 to CABG. There were no significant differences in baseline characteristics between the 2 treatment groups. The protocol was approved by the institutional review boards at the participating institutions.
Of the 1829 patients who were initially randomized in the BARI trial, 212 died or were inactivated during the next 4.5 years. The remaining 1617 patients were alive at 4.5 years and eligible for measurement of EF by RVG, which by protocol was to be performed between 4.5 and 5.5 years after study entry. Of those eligible, EF was measured in 1220, or 75%. The remaining 25% of patients did not undergo RVG for a variety of reasons, most commonly patient refusal, geographic inaccessibility, severe noncardiac illness, or patient death between 4.5 and 5.5 years after study entry.
Definitions used in the BARI trial have been published previously.3 4 8 Baseline angiographic data, including a measure of EF, were analyzed at the clinical sites and at the BARI angiographic core laboratory at Stanford University; analyses in this study were based on core laboratory angiographic data. Normal LV function was defined as EF ≥50, or when EF was missing, wall motion score ≤10. For this study, intercurrent MI was one occurring between randomization and 5-year RVG.
RVG was performed between 4.5 and 5.5 years after patient enrollment by the gated equilibrium technique according to detailed instructions provided in the study protocol. Acquisition and processing of the studies were performed by each center to determine EF. Because of budgetary constraints, no core laboratory processing was performed. A quality control study was completed by each center before its certification to perform RVG within the protocol.
Quality Control Study
As part of the quality control study, 20 patient studies were submitted to each center for processing by designated operators. All centers were then compared with respect to EF, end-diastolic frame and counts, end-systolic frame and counts, and background counts. The EF for each patient for each operator was compared with the mean for all operators to compute a correlation coefficient and a mean deviation. The median value among operators for the correlation coefficient was 0.97, and the median value for the deviation was 3.7 EF points. Prespecified criteria were used to identify and investigate quality control problems at each site on the basis of these data before site certification; there were few such problems.
Differences in baseline characteristics between patients with 5-year EF data available versus those with missing measurements and between patients with 5-year EF data randomized to PTCA versus CABG were assessed by χ2 tests or Fisher’s exact tests for dichotomous variables and t tests or Wilcoxon tests for continuous variables.
Differences in measured EF by treatment assignment in the entire cohort and in patient subgroups were assessed by t tests. Assuming an SD of 12 points for EF measurements, the power to detect a difference in EF of 2 points was 83%, and the power to detect a difference of 3 points was 99%.
Two secondary analyses were performed. In the first, a χ2 test was performed to compare the number of patients by treatment assignment who had a measured EF above or below 2 arbitrary cutpoints: 30 and 40. The second analysis was performed on the entire BARI randomized cohort to estimate the possible effect of missing values. Patients who were alive at 5.5 years but missing EF data were assigned an EF of 55.7 (the mean EF for patients with EF measurements). Patients who were dead at 5.5 years and missing EF data were assigned an EF of 0. Differences in measured-imputed EF were assessed by a Wilcoxon test.
A multivariate linear regression model was developed to predict 5-year EF from baseline clinical variables. Standard stepwise regression methods (with a P<0.10 entry criterion) were used to initially identify baseline factors associated with 5-year EF. A final model was created by forcing the treatment variable (PTCA versus CABG) into the initial model and then retaining variables with P<0.05. The parameter estimate in the linear regression model is the estimated change in EF given the category indicated compared with its complement (ie, female to male). The baseline LV function used in this model was measured by contrast ventriculography.
Selection of Patients
Among the patients eligible at 4.5 years to undergo RVG, the patients who underwent RVG were compared with those who did not. The 2 groups differed significantly with respect to a number of baseline characteristics (Table 1⇓). The patients in whom data from RVG were missing were as a group older, less likely to be white, and more likely to be female and to have hypertension, peripheral vascular disease, treated diabetes, unstable angina/non–Q-wave MI, and major ECG abnormalities.
The selection factors that influence the performance of RVG, however, did not appear to be preferentially associated with either PTCA or CABG. Table 2⇓ shows the baseline characteristics for 623 patients who underwent RVG after randomization to CABG compared with the 597 patients who underwent RVG after randomization to PTCA. These 2 groups did not differ significantly with respect to multiple baseline characteristics, demonstrating that the initial randomization process remained valid in the subgroup of patients who underwent RVG.
EF Measurements, Overall and by Initial Surgical Graft Method
For the total study group of 1220 patients, the EF in the CABG group was 55.8±12.3, compared with 55.7±12.7 in the PTCA group (P=0.82; Figure 1⇓). There was also no difference between the CABG group and the PTCA group in the prevalence of patients with low EFs (3.1% versus 3.4% for EF<30, P=0.77, and 10.0% versus 11.7% for EF<40, P=0.32).
When the analysis was restricted to patients who received their assigned treatment (n=1204) and the patients randomized to CABG were divided into those who received saphenous vein grafts only (n=109; EF 56.1±11.6) and those who received ≥1 internal mammary artery graft (n=505; EF 55.7±12.4), there was still no effect of initial revascularization on EF (P=0.96 for a 3-group ANOVA including those randomized to PTCA; n=591; EF 55.7±12.7).
Effect of Subsequent Surgery in the PTCA Group
Among patients who received their assigned treatment, patients randomized to PTCA who did not receive subsequent CABG had a higher EF at 5 years (n=414; EF 56.9±11.9) than patients who were randomized to PTCA and received subsequent CABG (n=177; EF 52.9±14.0) or those randomized to CABG (n=614; EF 55.8±12.2) (Figure 2⇓; P=0.0014 for a 3-group ANOVA and P=0.017 for PTCA only versus patients ever receiving CABG).
There was no significant difference in measured EF between patients randomized to PTCA and patients randomized to CABG within multiple subgroups (Table 3⇓; Figure 3⇓), including treated diabetes, no treated diabetes, 3-vessel disease, 2-vessel disease, complete revascularization, incomplete revascularization, prior or intercurrent MI, and no prior or intercurrent MI.
Linear Regression Models
The multiple linear regression model developed to predict 5-year EF by RVG from baseline parameters included the baseline LV function, which was determined by contrast ventriculography in most, but not all, patients (Table 4⇓). The 5-year RVG EF was higher in the presence of normal LV function at baseline and female sex. Five-year EF was lower in the presence of a major Q-wave on ECG, a history of prior MI, ≥1 diffuse coronary lesion, or ≥1 total coronary occlusion. Treatment assignment to PTCA or CABG was not significant (P=0.95).
Secondary Imputation Analysis
When EF values were assigned to patients with missing data, there was no evidence of a treatment difference regarding EF in the BARI randomized cohort (P=0.25, Table 5⇓). A difference favoring the CABG group was significant in the diabetic patients (P=0.0018) but not in the nondiabetic patients (P=0.83), reflecting the greater number of deaths in diabetic patients in the PTCA group.
These data demonstrate that in the BARI randomized trial, initial treatment assignment to angioplasty did not result in any difference in LV function compared with initial treatment assignment to surgery in patients alive at 5 years. Although there was considerable variability in LVEF measured at 5 years by RVG in both the surgery and angioplasty groups, the 2 distributions were very similar, as confirmed by the value of P=0.82 for the comparison between the 2 groups and the similar percentages of patients with low EFs.
There was also no difference in EF between patients randomized to surgery and those randomized to angioplasty in the designated subgroups of patients with and without complete revascularization, 3-vessel disease, treated diabetes, and prior MI. The use of an internal mammary artery graft at the time of surgery had no significant effect on EF. The multivariate model to predict 5-year EF on the basis of baseline parameters, including baseline ventricular function measurements, failed to show any significant effect of initial treatment assignment after adjustment for important baseline variables. Thus, in patients alive at 5 years in the BARI randomized trial who underwent RVG to determine EF, there is no evidence of any effect of initial treatment assignment. Although EF was higher in patients randomized to PTCA who did not undergo subsequent CABG, these patients were presumably those with the least eventful clinical course.
Ventricular function is the most important determinant of prognosis in patients with chronic coronary artery disease.5 9 10 11 Although its importance is generally acknowledged, the relative magnitude of its effect is not always recognized. For example, in medically treated patients in the CASS trial, the χ2 value associated with the angiographic LV score was 429, compared with a χ2 value of 72 for the number of diseased vessels, which was the next most important variable.11 Thus, ventricular function appeared to be prognostically much more important than the next most important variable. Similar results have been demonstrated in multiple other data banks.10 12 Several analyses have also confirmed the importance of ventricular function in determining prognosis after coronary revascularization.13 14 Although the significance of the extent of coronary artery disease is substantially reduced by coronary revascularization, baseline ventricular function remains important as a determinant of outcome after revascularization. Because of its recognized importance on the basis of all this evidence, LVEF measured at 5 years was prospectively designated as an important secondary end point at the time of the design of the BARI trial.
The primary end point of the BARI randomized trial was death and MI at 5 years after randomization.3 There was no overall difference in this end point according to initial treatment assignment, although a significant difference emerged with longer follow-up.6 Subsequent reports from the BARI randomized trial have included reports on cardiac mortality,7 diabetes,15 functional status,16 sex effects,17 and age.18 This article is the first to report the long-term ventricular function data.
Other randomized trials have been conducted to compare angioplasty and surgery as initial revascularization strategies.1 2 Data have been reported on clinical end points, treadmill exercise testing, exercise thallium imaging, and recurrent revascularization. We are not aware of any previous reports on long-term ventricular function from any of the other randomized trials comparing surgery and angioplasty.
This study has a number of limitations. Because of budgetary constraints, there was no core or central laboratory for the interpretation of the BARI ventricular function data. As outlined previously, an extensive quality control effort was undertaken to certify the Nuclear Cardiology Laboratory of each center participating in this study before the acquisition of RVG data. A set of standardized instructions for the performance of these studies was provided. There was generally excellent agreement between the centers on a sample group of 20 patient studies that were processed at each center. Prespecified criteria were used to identify and address possible quality control problems at individual centers. Despite these extensive efforts, however, the use of individual center processing probably increased the variability in the data compared with core laboratory processing. This increase in variability would reduce the power of this study to detect very small differences between the groups.
A second limitation is the absence of RVG data on 25% of the patients who were alive and eligible at 5 years. The patients who did not undergo RVG differed significantly from those who did with respect to a number of important baseline characteristics. They were older and more likely to have hypertension, peripheral vascular disease, treated diabetes, unstable angina or non–Q-wave MI, and major ECG abnormalities at the time of enrollment. These differences suggest that those patients who did not undergo RVG were “sicker” than those who did. It is therefore possible that this group had lower EFs than the patients who had measurements performed. Among those patients who underwent RVG, however, the patients randomized to surgery and the patients randomized to angioplasty were well matched with respect to multiple baseline characteristics. This would suggest that the selection process determining which patients did not undergo RVG was not associated with initial treatment assignment. The results presented here, ie, the absence of a difference according to initial treatment assignment, can probably be extrapolated to those patients who did not undergo RVG.
A third limitation is that an analysis of measurements performed at 5 years does not include any patients who died before that time. These results therefore apply only to patients alive at 5 years. This qualification is especially important with respect to the subgroup of patients with treated diabetes, in whom earlier reports from the BARI randomized trial showed differences in event rate with respect to initial treatment assignment.6 7 As shown in the secondary analysis, which used imputation, there were no overall differences between the CABG group and the PTCA group when patients who were dead and missing EF data were assigned an EF of 0. A difference was observed, however, in the subgroup of diabetic patients. The power of this study to detect small differences in some of the smaller subgroups of survivors should also be recognized. For example, only 177 patients in the subgroup with treated diabetes underwent RVG at 5 years. The power to detect a difference in EF of 2 points according to initial treatment assignment in this subgroup would be only 19%, and the power to detect a difference of 5 points would be 76%.
Finally, the number of patients with EF<50% at the time of randomization was modest (169/909, or 19%, with measured EF), and the number of patients with EF<30% was even smaller (7/909, or 0.8%). The power of this study to compare angioplasty and surgery with respect to the recovery of ventricular function in patients with chronically ischemic myocardium is very limited.
Despite these limitations, we believe that these data prove that in patients alive at 5 years, initial treatment assignment to angioplasty is not associated with a difference in long-term ventricular function compared with initial treatment assignment to surgery. These data extend the conclusions of the previous publications from the BARI randomized trial. In patients with multivessel coronary disease who are eligible for treatment with either angioplasty or surgery, initial treatment with angioplasty is not associated with a penalty with respect to long-term ventricular function among 5-year survivors of coronary revascularization.
- Received May 24, 2000.
- Revision received October 27, 2000.
- Accepted October 27, 2000.
- Copyright © 2001 by American Heart Association
Mock MB, Ringqvist I, Fisher LD, et al. Survival of medically treated patients in the Coronary Artery Surgery Study (CASS) registry. Circulation. 1982;66:562–568.
Chaitman BR, Rosen AD, Williams DO, et al, on behalf of the BARI Investigators. Myocardial infarction and cardiac mortality in the Bypass Angioplasty Revascularization Investigation (BARI) randomized trial. Circulation. 1997;96:2162–2170.
Emond M, Mock MB, Davis KB, et al. Long-term survival of medically treated patients in the Coronary Artery Surgery Study (CASS) Registry. Circulation. 1994;90:2645–2657.
Harris PJ, Harrell FE Jr, Lee KL, et al. Survival in medically treated coronary artery disease. Circulation. 1979;60:1259–1269.
Hammermeister KE, DeRouen TA, Dodge HT. Variables predictive of survival in patients with coronary disease: selection by univariate and multivariate analyses from the clinical, electrocardiographic, exercise, arteriographic, and quantitative angiographic evaluations. Circulation. 1979;59:421–430.
BARI Investigators. Influence of diabetes on 5-year mortality and morbidity in a randomized trial comparing CABG and PTCA in patients with multivessel disease: the Bypass Angioplasty Revascularization Investigation (BARI). Circulation. 1997;96:1761–1769.
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.
Jacobs AK, Kelsey SF, Brooks MM, et al. Better outcome for women compared with men undergoing coronary revascularization: a report from the Bypass Angioplasty Revascularization Investigation (BARI). Circulation. 1998;98:1279–1285.