Author + information
- Received May 13, 1996
- Revision received October 15, 1996
- Accepted October 18, 1996
- Published online February 1, 1997.
- Barbara E Tardiff, MDA,*,
- Robert M Califf, MD, FACCA,
- Douglas Morris, MD, FACCB,
- Eric Bates, MD, FACCC,
- Lynn H Woodlief, MSA,
- Kerry L Lee, PhDA,
- Cindy Green, MSA,
- Wolfgang Rutsch, MDD,
- Amadeo Betriu, MDE,
- Philip E Aylward, MD, FACCF,
- Eric J Topol, MD, FACCG,
- for the GUSTO InvestigatorsA
- ↵*Dr. Barbara E. Tardiff, Box 3850, Duke University Medical Center, Durham, North Carolina 27710.
Objectives. This study sought to investigate the impact of surgical revascularization on outcome after myocardial infarction.
Background. Small variations in rates of coronary artery bypass graft surgery (CABG) were noted among thrombolytic regimens in the Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries (GUSTO) trial, prompting the question of whether survival differences were partly related to differences in CABG rates.
Methods. Patients in the GUSTO trial were randomized to one of four thrombolytic strategies. Of 40,861 patients with complete data, 3,526 underwent surgical revascularization during their initial hospital admission. Thirty-day and 1-year mortality rates were estimated using Kaplan-Meier techniques, and the impact of CABG as a time-dependent covariate on death was evaluated using a Cox survival model, adjusting for baseline prognostic factors.
Results. The median time from study enrollment to CABG was 7 days across treatment groups. A 15% reduction in mortality for the tissue-type plasminogen activator (t-PA)–treated group was evident by the seventh day. Bypass surgery was a significant independent predictor of 30-day mortality (risk ratio 1.87) and a weaker predictor of 1-year mortality (risk ratio 1.21). Operative mortality was highest in patients with acute mitral regurgitation, ventricular septal defect or poor left ventricular function and in those undergoing CABG within the first 4 days of randomization.
Conclusions. The survival benefit of accelerated t-PA was not related to surgical revascularization. Bypass surgery was associated with excess mortality in the first year, but the added short-term mortality associated with CABG may be balanced by anticipated long-term benefit in specific groups of patients.
(J Am Coll Cardiol 1997;29:240–9)
The Global Utilization of Streptokinase and Tissue Plasminogen Activator (t-PA) for Occluded Coronary Arteries (GUSTO) trial () demonstrated that accelerated administration of t-PA resulted in improved survival compared with streptokinase and that this survival advantage is related to earlier infarct-related artery patency and enhanced perfusion (). Initial reports from the trial () also noted small differences in rates of coronary artery bypass surgery (CABG) among treatment regimens and raised the question of whether survival differences were attributable to differences in operation rates.
The large population of the GUSTO trial (41,021 patients admitted to 1,081 institutions in 15 countries) offered the opportunity to examine relations between baseline clinical and angiographic findings and early surgical revascularization. We attempted to define the impact of surgical revascularization on mortality and other clinical outcomes and to differentiate between outcome effects attributable to CABG procedures and those related to the thrombolytic strategy and patient baseline characteristics.
1.1 Patient population.
Enrollment criteria for the GUSTO trial have been previously described (). The protocol was approved by the institutional review board at each study center and patient written informed consent was obtained before randomization. The 41,021 patients were enrolled between December 27, 1990 and February 22, 1993, and data from 40,861 patients were complete for inclusion in this study (99.6%). Data from the remaining GUSTO patients (0.4%) were incomplete or inadequate for the purposes of this study, and these patients were excluded from analyses. For the purposes of the present report, patients were classified into two groups: those undergoing CABG at any time during the initial hospital period and those not undergoing CABG.
1.2 Study protocol.
Patients were randomly assigned to one of four thrombolytic strategies: 1) 1.5 million U of streptokinase (SK) over a 60-min period with 12,500 U of subcutaneous (SQ) heparin twice daily beginning 4 h after the initiation of thrombolytic therapy; 2) 1.5 million U of SK with intravenous (IV) heparin administered as a bolus dose of 5,000 U followed by an infusion of 1,000 U/h; 3) t-PA given as a bolus dose of 15 mg followed by 0.75 mg/kg body weight over 30 min (maximal 50 mg) and then 0.5 mg/kg infused over 60 min (maximal 35 mg) with IV heparin administered as in strategy 2; or 4) combination therapy with 1.0 million U of SK administered over a 60-min period concomitantly with t-PA (1.0 mg/kg with 10% given as a bolus) along with IV heparin.
Aspirin and atenolol were administered as specified by the study protocol (). All other medications and interventions, including cardiac catheterization and surgical or percutaneous revascularization, were at the discretion of the investigator.
1.3 End points.
The primary end point was death from any cause within 30 days of enrollment. The following additional end points were prospectively defined and assessed in-hospital: hemorrhagic or nonhemorrhagic stroke; death or nonfatal stroke; death or nonfatal hemorrhagic stroke; death or nonfatal disabling stroke; reinfarction; ischemia; shock; and congestive heart failure or pulmonary edema. Reinfarctionwas determined by the presence of at least two of following criteria: recurrent ischemia lasting >15 min; new ST-T wave changes or new Q waves; reelevation of cardiac enzyme levels above the normal limit or further elevation >20% above normal; or angiographic reocclusion of a previously patent infarct-related artery. Shockwas defined as a systolic blood pressure <90 mm Hg unresponsive to fluid therapy associated with signs of hypoperfusion or cardiac index ≤2.2 liters/min per m2. Strokewas characterized as new neurologic deficits lasting >24 h or resulting in death. Congestive heart failurewas defined as signs or symptoms of congestion or low cardiac output. Recurrent ischemiawas defined as symptoms, electrocardiographic changes or new hypotension, pulmonary edema or murmur thought to represent myocardial ischemia. The date of the first episode of each end point was recorded.
1.4 Data management and statistical analysis.
Demographic and clinical data were collected on standardized case report forms. Definitions of variables and written instructions were provided for completion of forms. Quality checks were conducted regularly and missing or inconsistent data values were investigated. Data from a randomly derived and representative sample (12%) of forms were verified from the hospital medical record.
Baseline characteristics of the surgical and nonsurgical treatment groups were compared for categoric data by summarizing frequencies and percentages and for continuous data as mean value ± SD or median (25th and 75th percentiles).
Binary outcome variables were similarly reported as percentages for in-hospital events. Kaplan-Meier techniques were applied to estimate the rate of CABG from 7 to 30 days of enrollment. The raw event rates included a number of medically treated patients who died early, before having the opportunity for surgical therapy. To adjust for bias resulting from the inclusion of the early deaths in the nonsurgical treatment group, outcome event rates were also calculated after subtracting from each group those patients who died or who had experienced the given outcome event before specific time points.
Mortality risk predictions were calculated on the basis of a previously described survival model that includes clinical characteristics measured at the time of entry into the trial and found to be predictive of mortality (). The distribution of predicted risk across the patient groups was illustrated by plotting the cumulative distribution of predicted risk of 30-day death for the patients treated with and without CABG.
Mortality during the 30-day and 1-year follow-up periods was depicted using Kaplan-Meier curves. The impact of CABG on time until death after acute myocardial infarction (MI) was further evaluated by construction of a Cox survival model, adjusting for known baseline prognostic factors and examining the effect of operation as a time-dependent covariate. In this analysis, patients eventually treated surgically were considered to be members of the medically treated group until the day of operation. At that point, they were crossed over to the surgically treated group.
Multivariable regression analysis was used to construct a model to examine the relation between baseline clinical characteristics and death within 30 days of randomization for patients who underwent a CABG during the hospital period. Characteristics of the variables considered for the model were reported as frequencies and percentages for categoric data, and the median (25th, 75th percentiles) percentiles were used for continuous variables. Sixty-five studies that were missing the end point or all catheterization data were excluded. All remaining 3,461 studies were used in the logistic regression model. A method for simultaneous imputation of predictor variables based on the concepts of maximal generalized variance and canonical variables was used to estimate missing predictor variables to allow for the analysis of all patients. The variables were analyzed to assess the assumption that the log odds of death is linearly related to the dependent variables, and adjustments were made when necessary. The receiver operating characteristic curve was used to show model performance. This index measures the concordance between predictive values and actual outcomes. The model was calibrated using a graph of the average predicted values versus the observed mortality rates across deciles of risk.
2.1 Patient characteristics.
Nine percent of patients treated with t-PA and 8.3% of those treated with SK (IV and SQ heparin treatment groups combined) underwent surgical revascularization during the initial hospital period (p = 0.050). Baseline demographic characteristics and presenting features of patients referred for CABG were comparable to those treated nonsurgically, although there were small differences in individual variables (Table 1). Overall, patients who underwent CABG were slightly more likely to have features consistent with long-standing cardiovascular disease (history of angina or previous MI; comorbid disorders associated with cardiovascular disease, including hypertension, diabetes mellitus and hypercholesterolemia; or a positive family history). Fig. 1shows a summary of mortality risk predictions for the surgical and nonsurgical treatment groups. Because the nonsurgically treated group included more very low risk and extremely high risk patients, analysis of overall risk factor characteristics led to a higher mean predicted (7.1% vs. 6.6%) but a lower median predicted (3.3% vs. 3.9%) 30-day mortality.
Catheterization results were available for 22,640 patients (55% of the total) and are included in Table 2. Patients referred for CABG had more extensive disease and a higher rate of multivessel involvement than those treated nonsurgically (p < 0.001). The patients treated with CABG were similar to the nonsurgical population in terms of relative distribution of infarct-related vessels and patency of the infarct-related artery at the time of first catheterization but tended to have a slightly lower median left ventricular ejection fraction (50%) than the nonsurgical group (52%) (p < 0.001).
2.2 Timing of surgical intervention.
The time from randomization to CABG for each treatment group is shown in Table 3. There were no significant differences in the timing of CABG.
Thirty-day mortality as a function of time between randomization and CABG is displayed in Fig. 2. Mortality was substantial in patients undergoing CABG within the first 2 days but decreased rapidly after that. There was no apparent relation between time from onset of symptoms to CABG and mortality beyond the third day.
Fig. 3illustrates the mortality data for each of the treatment groups compared with the occurrence of CABG procedures. A statistically significant difference in mortality was observed between the accelerated t-PA group and the two SK regimens, as has been previously reported (). A reduction in mortality for the t-PA–treated group is evident within 24 h of randomization. The reduction in mortality in the t-PA group becomes more substantial over the first week after study entry and persists through the 30-day follow-up period. Patients underwent CABG procedures at a similar rate during the first week after study enrollment, regardless of treatment group assignment. There was a small excess of CABG procedures in the t-PA–treated group from 1 week until the 30-day end point.
2.3 Thirty-day clinical outcomes.
Unadjusted outcome event rates in the CABG and non-CABG groups are presented in Table 4. Patients with acute MI complicated by ventricular arrhythmias, hypotension and mechanical disturbances, such as mitral regurgitation and acute ventricular septal defects, comprised a slightly higher percentage of the CABG group than the non-CABG group (p < 0.001 for all). The overall incidence of stroke was similar in the two groups (p = 0.188), although the CABG group was more likely to have a nonhemorrhagic stroke. Nearly all the recurrent ischemia and reinfarction events antedated the surgical procedure (Table 5). Congestive heart failure and pulmonary edema occurred principally in the preoperative and immediate perioperative periods. The timing of complications with respect to the surgical procedure is not known for outcomes reported in Table 4but not included in Table 5.
On the basis of these unadjusted outcome data, patients undergoing CABG appeared significantly less likely than those not undergoing CABG to experience a fatal event in the 30-day follow-up period. However, as evident from Fig. 3, a significant proportion of the deaths in all groups occurred within the first 24 h. Outcome event rates adjusted to eliminate patients who died before 24, 48, 72 and 96 h after onset of symptoms are shown in Table 6. When patients who died within the first 4 days of onset of symptoms were not included in the analyses, CABG was associated with a higher 30-day mortality than nonsurgical treatment.
Coronary artery bypass graft surgery was a predictor of 30-day mortality when examined as a time-dependent covariate in a Cox survival model (chi-square value 53.0, p < 0.001). Without adjusting for other prognostic factors, the likelihood of death increased by a factor of 1.94 (1.64, 2.28) with the occurrence of CABG (beta value 0.6615). When the model was adjusted to compensate for differences in baseline characteristics that have been found to be predictive of death (), CABG remained an independent predictor of 30-day mortality (chi-square value 49, p < 0.001). The risk ratio was 1.87 (1.59, 2.20) in this adjusted model (CABG beta value 0.627).
2.4 One-year mortality outcomes.
Fig. 4illustrates survival curves for the surgical and nonsurgical groups plotted over the period from randomization to 1 year. An excess early death rate is evident in both groups. A third plot shows survival over the period from 8 days to 1 year in the nonsurgically treated patients who were alive at 8 days, the mean time of CABG. From 30 days to 1 year, the curves parallel each other and begin to converge.
The Cox model was also extended to include 1-year mortality outcome data. Bypass surgery remained an independent but weaker predictor of 1-year mortality (unadjusted chi-square value 13, p < 0.001; adjusted chi-square value 7, p < 0.006) and was associated with less excess risk (unadjusted beta value 0.2500, risk ratio 1.28 [1.12, 1.47]; adjusted beta value 0.1868, risk ratio 1.21 [1.06, 1.38]).
2.5 Risks factors for operative mortality.
In a multivariable analysis, the risk of death after CABG and within 30 days of enrollment was directly related to the development of an acute ventricular septal defect or acute mitral regurgitation and inversely related to left ventricular ejection fraction and the time from enrollment until operation (Table 7). A substantial decrease in mortality rate was seen as time until operation approached 4 days; a slight decrease was seen thereafter. The effects of these variables were independent of and in addition to the influence of clinical characteristics measured at the time of study entry and previously found to be predictive of mortality in the GUSTO population (the logit GUSTO). The infarct-related artery and number of diseased vessels were not significant predictors of mortality, nor were regional differences in perioperative mortality observed.
3.1 Bypass surgery and thrombolytic strategy.
This analysis of the GUSTO trial data base confirms that the observed benefit of accelerated t-PA was not due to the increased use of revascularization. An in-depth examination of the data demonstrated that the slightly increased frequency of CABG in the t-PA group did not explain the mortality differences observed between the t-PA– and SK-treated groups. A 15% statistically significant reduction in mortality afforded by t-PA was apparent within the first week of enrollment, during which time the proportion of patients undergoing CABG was not different among the four treatment groups. The early divergence of the survival curves eliminates surgical revascularization as an explanation for the survival benefit of accelerated t-PA.
A slightly higher rate of CABG in the t-PA–treated patients was observed later in the study period. This higher procedure rate may be attributed to the survival advantage of t-PA and larger number of surviving patients available for referral. Patients treated with t-PA had better preservation of left ventricular function that those treated with SK (), which may have also improved their likelihood of referral for CABG. We cannot exclude a role for surgical revascularization as an important part of the strategy for preserving myocardial function after thrombolysis with t-PA.
3.2 Clinical outcomes.
Excess mortality in the first year after operation was observed for patients treated with CABG during the index hospital period after acute MI. The 30-day mortality rate for the nonsurgical group was lower than that for patients undergoing CABG when deaths that occurred before the time that surgical revascularization would normally be considered were eliminated from the analysis. When only deaths occurring after 96 h were considered, the mortality rates were 3.1% for the surgically treated group versus 2.1% for patients not surgically treated. These findings are comparable to previously reported mortality rates for CABG in the setting of acute MI ([5, 6]).
Coronary artery bypass graft surgery was an independent predictor of mortality at 30 days and at 1 year in a time-dependent covariate model adjusted for baseline characteristics that have previously been shown () to be predictive of mortality. A recently published systematic overview of randomized trials of CABG versus medical therapy demonstrated () that in predominantly elective CABG, the benefit does not become evident until after the first year of randomization. At 1 year after randomization to CABG, the rate of infarction or death reported in the overview was 11.6% in the CABG group and 8.0% in the medically treated group. A similar pattern was observed in the Duke experience ([8, 9]). As in the GUSTO trial, those studies found a small but consistent excess mortality in the perioperative period, with survival curves converging over the first year, crossing just after the first year and then showing a benefit of operation after that time. The results presented in this study should not be interpreted as indicating that the surgical choice in these patients was incorrect; rather, the full impact of treatment selection will be discernible only with longer term follow-up.
Risk factors for early mortality after CABG after thrombolysis include mechanical complications of acute MI (ventricular septal defect or acute mitral regurgitation), poor systolic function (decreased left ventricular ejection fraction) and operation within the first few days of symptom onset. The appropriate timing of operation after acute MI is a complex issue. These and other recent data () support the concept that the myocardium benefits from an initial period of stabilization. However, patients referred for early operation were most likely to have experienced acute deterioration immediately before CABG, and it is difficult to fully account for all factors contributing to the higher risk associated with operating on a clinically unstable patient. Although the operative mortality rate of patients undergoing CABG within the first several days of acute MI is higher than in those who had CABG later in their hospital course, it is not clear that the mortality rate of these critically ill patients would be lower if CABG were deferred or not performed. Indeed, although patients with acute ventricular septal defect were at high risk for operative mortality, patients treated surgically fared better (30-day mortality rate 50%; 1-year mortality rate 53%) than those managed medically (30-day mortality rate 79%; 1-year mortality rate 83%) (). Patients with left ventricular dysfunction also have a high operative mortality rate but enjoy a substantial survival benefit with CABG ().
A higher incidence of nonfatal events, including recurrent ischemia and infarction, shock and congestive heart failure, was seen in the surgically revascularized group. However, the majority of events occurred before the time of CABG and thus constituted indications for CABG rather than consequences of the operation. Patients undergoing CABG had slightly higher overall baseline morbidity as well as median predicted mortality on the basis of factors found to be predictive of survival. However the mean predicted mortality was lower than that of the nonsurgical group, reflecting the fact that very elderly or profoundly hemodynamically compromised patients were very unlikely to be selected for CABG.
A total of 8,519 patients in the nonsurgical group and 386 patients in the surgical group underwent percutaneous revascularization. Because of the complexity of this issue and the variety of circumstances under which percutaneous revascularization was used, and the primary importance of examining the effect of CABG on the treatment effect of thrombolytic therapy, the details of outcome in the patients undergoing percutaneous revascularization will be included in a separate report. Including or excluding these patients from the nonsurgical group made no substantive difference in the estimate of excess risk in the surgical group in the first year.
3.3 Limitations of the study.
There are several potential sources of bias in this analysis. Although patients were randomized to the various thrombolytic treatments, surgical intervention was at the discretion of the clinician. The amount of data that could be realistically collected by the case report forms precluded a comprehensive understanding of what factors predisposed patients to be referred for CABG after admission or after attempted percutaneous revascularization. During the first few days of enrollment, decisions as to need for and timing of operation were most likely based on the severity of clinical symptoms; that is, patients who did not respond to thrombolytic therapy may have been directed toward early surgical intervention. Later in the index hospital period, CABG may have been indicated by more extensive or generalized disease.
An attempt was made to adjust for differences in baseline characteristics that have been found to be predictive of death, but statistical adjustment cannot compensate completely for the differences in the acuity and severity of illness between the surgical and nonsurgical groups. In particular, complete data regarding coronary anatomy and left ventricular function were not available for all patients, and therefore it was not possible to adjust for differences in these clinical characteristics. It is possible that the significance of CABG in the time-dependent covariate model would have been diminished if left ventricular function and the severity and distribution of coronary disease were taken into account. Details regarding additional surgical procedures, valve repair or replacement, aneurysm resection or carotid endarterectomy, often performed at the time of revascularization, are also limited. Associated procedures may have increased the operative risk and affected long-term outcome.
Another obvious bias is that some patients may not survive long enough to undergo CABG, and all these patients by default are medically treated. It is difficult to account for the effect of these early deaths when trying to understand the impact of CABG on outcome. We attempted to compensate for this influence in two ways: 1) by eliminating patients from the analysis who died shortly after enrollment, and 2) by constructing the time-dependent covariate model (which included all patients). Results from these two approaches were similar, and both showed an increased relative risk for 30-day mortality for patients undergoing surgical revascularization compared with those treated nonsurgically.
Finally, the excess mortality observed in patients undergoing CABG can be interpreted as an association, but causality cannot be assumed or proved. Patients requiring surgical revascularization after acute MI and thrombolysis were at increased risk for mortality at 30 days and at 1 year, but this analysis cannot completely separate excess mortality resulting from complications related to the surgical procedure from the added disease burden prompting early surgical intervention.
3.4 Clinical implications.
Although thrombolytic therapy results in improved survival in patients with acute MI, it does not cure or slow the progression of atherosclerotic heart disease. Untreated individuals remain at high risk for recurrent ischemic events and infarction, with significant ischemia documented in 10% to 35% of patients and reinfarction in 2.5% to 5.4% ([13–17]). Patients experiencing recurrent ischemic events are at increased risk for hypotension, congestive heart failure, conduction abnormalities and death (). Many patients can be managed medically, but a substantial number are ultimately referred for percutaneous or surgical revascularization. A number of studies have examined acute and long-term outcome in patients with acute MI undergoing CABG, but there are limited randomized data from the prethrombolytic era and no randomized studies evaluating the safety and efficacy of CABG in patients after thrombolysis.
Previously published investigations report in-hospital mortality rates for CABG after acute MI, ranging from 2.6% () to 6.7% ([20, 21]). A recent retrospective analysis of 2,175 patients undergoing CABG at a California health maintenance organization from January 1990 through April 1993 failed to show any increase in mortality and morbidity for the 530 patients who underwent CABG within 1 month of acute MI, with operative mortality ranging from a low of 0% in 30 patients operated on between 24 and 72 h to a high of 4.4% in 23 patients operated on within 24 h of acute MI (). The operative mortality rate in 1,645 patients without a recent infarction was 1.9%. In a small randomized trial () comparing CABG with medical treatment in 68 patients presenting within 4 h of acute MI, the 3-month mortality rate for patients treated with CABG was 2.9%. Medical treatment alone, which did not include thrombolytic therapy, was associated with a 20.6% mortality rate.
These data confirm those of other investigators that surgical revascularization is accomplished with acceptable mortality after acute MI. Although complications associated with the perioperative period may negatively impact overall short-term mortality rates, many patients experience substantial benefit from early augmentation of myocardial perfusion. Better delineation of the specific subpopulations that stand to improve the most from costly strategies, such as CABG, will be necessary as strategies for health care cost containment are reevaluated.
☆ This study was funded by a combined grant from Bayer, New York, New York; CIBA-Corning, Medfield, Massachusetts; Genentech, South San Francisco, California; ICI Pharmaceuticals, Wilmington, Delaware; and Sanofi Pharmaceuticals, Paris, France. A complete list of the GUSTO investigators appears in reference .
- coronary artery bypass graft surgery
- myocardial infarction
- tissue-type plasminogen activator
- Received May 13, 1996.
- Revision received October 15, 1996.
- Accepted October 18, 1996.
- The American College of Cardiology
- for the GUSTO-I Investigators,
- Lee KL,
- Woodlief LH,
- Topol EJ,
- et al.
- and the TAMI Study Group,
- Kereiakes DJ,
- Topol EJ,
- George BS,
- et al.
- Moosvi AR,
- Khaja F,
- Villanueva L,
- Gheorghiade M,
- Douthat L,
- Goldstein S
- Mark DB,
- Nelson CL,
- Califf RM,
- et al.
- Braxton JH,
- Hammond GL,
- Letsou GV,
- et al.
- for the GUSTO-I Investigators,
- Crenshaw BS,
- Morris DA,
- Pieper KS,
- et al.
- Parisi AF,
- Khuri S,
- Deupree RH,
- Sharma GV,
- Scott SM,
- Luchi RJ
- Chesebro JH,
- Knatterud G,
- Roberts R,
- et al.
- for the TAMI Study Group,
- Califf RM,
- Topol EJ,
- Stack RS,
- et al.
- for the Tami Study Group,
- Ohman EM,
- Califf RM,
- Topol EJ,
- et al.
- DeWood MA,
- Notske RN,
- Berg R Jr.,
- et al.
- Koshal A,
- Beanlands DS,
- Davies RA,
- Nair RC,
- Keon WJ