Prediction of 1-Year Survival After Thrombolysis for Acute Myocardial Infarction in the Global Utilization of Streptokinase and TPA for Occluded Coronary Arteries Trial
Background—When a patient survives thrombolysis for acute myocardial infarction, little information from large studies exists from which to estimate prognosis during follow-up visits.
Methods and Results—Baseline, in-hospital, and later survival data were collected from 41 021 patients enrolled in Global Utilization of Streptokinase and TPA for Occluded Coronary Arteries, a randomized trial of 4 thrombolytic-heparin regimens with standard aspirin and β-blockade. Cox proportional hazards models were developed to predict 1-year survival in 30-day survivors (n=37 869) from baseline clinical and ECG factors and in-hospital factors; a combined model then was developed (C-index 0.800). The model was simplified into a nomogram to predict individual outcomes (C-index 0.754). Factors reflecting demographics (advanced age, lighter weight), larger infarctions (higher Killip class, lower blood pressure, faster heart rate, longer QRS duration), cardiac risk (smoking, hypertension, prior cerebrovascular disease), and arrhythmia were important predictors of death between 30 days and 1 year. Black race was associated with a substantial increase in risk after considering other factors. Revascularization was associated with reduced risk between 30 days and 1 year.
Conclusions—When evaluating a patient who has survived acute infarction treated with thrombolysis, clinicians can estimate the likelihood of survival from factors easily measured during admission. Although many risk factors clearly relate to age, left ventricular dysfunction, or clinical instability, black race is an unexplained risk factor requiring further examination.
Although much is known about the prognosis of patients with acute myocardial infarction (MI), it is the major cause of death worldwide1 and seems likely to remain so well into the next millennium. Given the number of patients affected, the need to develop more precise estimates of their prognosis remains an important goal.
When patients with symptoms of MI are first evaluated, clinicians make decisions based on the history, physical examination, and ECG. Fibrinolysis is now accepted as standard therapy2 except when angioplasty is immediately available.3 During hospitalization, further decisions are made about acute care based on the clinical course. After the acute phase, the patient, family, and clinician meld data from previous history, new events, and hospitalization to form an understanding of future expectations and to plan interventions (including behavioral changes, medication, and revascularization) to improve outcome.
The Global Utilization of Streptokinase and TPA (alteplase) for Occluded Coronary Arteries (GUSTO-I) trial4 presents a unique opportunity to consider critical aspects of assessment after the acute phase in patients with ST-segment elevation MI eligible for fibrinolysis. Baseline characteristics and clinical outcomes were measured in all patients, a diverse population that spanned the spectrum of risk and interventions. We have described the relations between baseline clinical characteristics and 30-day outcomes5 6 and between baseline clinical and ECG findings and these outcomes.7 Our current goal was to include factors from the hospital course in a model that would allow patients and physicians to calculate a robust prediction of mortality risk after the acute phase (30 days) through 1 year after first assessment.
The GUSTO-I trial enrolled 41 021 patients with acute MI who presented with ≥1 mm ST-segment elevation in ≥2 limb leads or ≥2 mm in ≥2 contiguous precordial leads within 20 minutes to 6 hours after symptom onset.4 Exclusion criteria related to the risk of bleeding or allergy and not to age, cardiogenic shock on arrival, or prior bypass surgery.
Patients were randomly assigned to receive 1.5 MU streptokinase over 1 hour, with 12 500 U subcutaneous heparin twice daily starting 4 hours after thrombolysis began; 1.5 MU streptokinase over 1 hour, with 5000 U intravenous heparin followed by 1000 U/h, with dosing adjusted for activated partial thromboplastin time; 15 mg alteplase, then infusion of 0.75 mg/kg (50 mg maximum) over 30 minutes and 0.5 mg/kg (35 mg maximum) over the next hour, with the same intravenous heparin regimen; or alteplase (1.0 mg/kg over 1 hour, 90 mg maximum, 10% given as a bolus) and streptokinase (1 MU over 1 hour), given simultaneously but separately, with the same intravenous heparin regimen.
Patients received ≥160 mg chewable aspirin on arrival, then 160 to 325 mg daily. Atenolol (10 mg IV) was given in 2 doses to patients without contraindications and 50 to 100 mg orally daily thereafter. Other medications and procedures were left to the discretion of the physician.
At enrollment, detailed personal data were collected, including addresses and telephone numbers of the patient and a contact. At 1 year, 30-day survivors were to return a postcard providing survival information.6 Patients and contacts were called if postcards were not returned. A locator service was used to find US patients if these methods failed. Because of the large enrollment in GUSTO-I, only survival status was obtained at 1 year. Final follow-up was 96% overall: 98% in the United States and 95% elsewhere. One-year survival data were received for 96% of each treatment group.
We analyzed only patients surviving 30 days after enrollment, including 918 (2.5%) inpatients. Survival estimates were based on postenrollment days 30 to 365, inclusive. Events occurring after day 365 were censored on day 365. For descriptive purposes, “survived to 1 year” describes patients who survived to day 365 and were censored after then as being alive. “Died between 30 days and 1 year” describes patients who survived 30 days but were known to have died before day 366.
Four models were developed with Cox proportional hazards techniques, each containing all significant variables from the 30-day clinical5 and clinical/ECG7 models as appropriate, and those available that we considered potential predictors. The models contained (1) baseline clinical characteristics, (2) baseline clinical and ECG characteristics, (3) both of these plus in-hospital factors, and (4) all of these plus angiographic factors. (Complete statistical methods are available at http://dcri.duke.edu.) For each set of variables, we first assessed whether they violated the assumption of proportional hazards; all factors retained in final models met this assumption.8 We then determined the univariable relation of each factor with 1-year survival by using appropriate spline transformations9 for continuous variables that showed nonlinear relations. We combined all factors in the set and used stepwise variable selection to find the subset that jointly contained significant prognostic information. Once done for each of the 4 sets of variables, we combined them and performed stepwise variable selection again to determine which of all factors jointly best predicted survival from 30 days to 1 year.
For the baseline clinical/ECG model and the model also including in-hospital factors, we show the relative prognostic importance of factors by plotting the Wald χ2 minus the degrees of freedom from the full models. For the full, final model, hazard ratios and 95% confidence intervals show the multivariable relation of each factor to 1-year mortality rate. Nomograms were created with S-Plus software and the NOMOGRAM function.10
Of the 37 869 patients who survived 30 days, 1084 (2.9%) died between 30 days and 1 year. Most baseline predictors of 30-day death remained important for later outcomes (Table 1⇓). Expected risk factors dominated among patients who died, including demographics (age, weight, sex), measures of infarct size (anterior MI, worse Killip class, lower blood pressure), and historical risk factors (smoking, hypertension, prior cerebrovascular disease, prior MI, prior bypass); the poor prognosis of black patients was an exception. The most important independent predictors were age and heart rate. Historical factors generally were less important than the clinical factors.
Of the baseline ECG descriptors (Table 2⇓), heart rate and QRS duration were the critical factors, as was ECG evidence of prior MI. These factors negated the independent importance of many of the conduction disturbances.
Almost all in-hospital complications were associated with higher mortality rates among patients who survived the first 30 days (Table 3⇓). Those reflecting heart failure dominated prognosis, whereas recurrent ischemic events were less powerful. Revascularization was associated with significantly better late survival.
The relations between angiographic variables and late death are shown in Table 4⇓. Ejection fraction was the significant predictor; the degree of left main stenosis was of borderline significance.
The independent, significant predictors of survival from the history, physical examination, initial ECG, and hospitalization are shown in Table 5⇓. Age remained the dominant determinant of outcome; in-hospital heart failure and arrhythmic complications also were strong predictors. Revascularization was associated with substantially improved late survival. The relations between selected continuous variables and late survival are displayed in Figure 1⇓. Increasing age conferred a constantly increasing risk, whereas heart rate showed a U-shaped relation with higher mortality rates at the extremes.
The model taking into account clinical, initial ECG, in-hospital, and angiographic variables is shown in Table 6⇓. Ejection fraction added substantial information, but number of diseased vessels did not, once other clinical factors were considered.
Data from Tables 5⇑ and 6⇑ were first converted into 2 algorithms (available at http://dcri.duke.edu). Because most prognostic data were contained in only a few variables, simpler nomograms were created without (Figure 2⇓) and with (Figure 3⇓) angiographic data. With these, physicians can more easily estimate the mortality risk for 30-day survivors of MI. The models had excellent ability to discriminate patients by risk of death, with respective C-index values of 0.800 and 0.805 for the complete models without and with angiographic data and values of 0.754 and 0.786 for the nomograms without and with angiographic data. A C-index value of 0.800 means that for 80% of pairs of patients, the model would predict longer survival for the patient who actually did survive longer.
Better understanding of the pathophysiology of acute coronary syndromes has led to a focus on treatment of the acute phase of MI. This effort has yielded a significant reduction in mortality rates. Although effective therapies exist for secondary prevention, little quantitative information is available for use in risk stratification after acute events. This study shows the complexity of the situation. Many types of pathophysiological factors are independently associated with mortality risk, and focusing on one aspect without attention to others is unlikely to modify underlying risk substantially.
Almost every study of prognosis after MI has identified age as a critical factor. The patient’s age is a far greater determinant of mortality risk than any features discernible at follow-up visits. Although this factor is unmodifiable, the increasing proportion of elderly patients with MI should spur research into the benefits and risks of more aggressive therapies for this population. Despite their markedly increased risk11 and higher rates of multivessel coronary disease and left ventricular dysfunction,12 however, older patients are much less likely than younger patients to undergo invasive procedures aimed at improving prognosis.13
Being lighter or shorter in stature has prognostic importance in acute coronary syndromes.5 The striking impact of race was greater than expected. Race was not prognostic in the short term in GUSTO-I,5 but black patients had a >2-fold higher risk of late death after adjustment for other prognostic factors. Many studies have pointed out the lesser use of revascularization among racial minorities in the United States,14 but we noted an adverse prognosis even after accounting for revascularization. Differential access to or use of secondary prevention measures should be a source of further research into race-related differences in outcomes.
Left Ventricular Function
Many features reflecting the extent of myocardial damage were important. Most prominent among these were in-hospital heart failure or frank pulmonary edema and tachycardia.15 16 These findings point to the importance of assessments of patients with transient heart failure to determine whether myocardial function is affected and to optimize treatment with agents such as ACE inhibitors17 and nitrates.18
Arrhythmic complications also were associated with increased late mortality rates, but we found no evidence of a difference in later risk with supraventricular versus ventricular arrhythmia. This is surprising because ventricular arrhythmias have been assumed to confer a worse prognosis than supraventricular arrhythmias. Both are associated with poor left ventricular function, however.19 20
Many guidelines recommend a measure of left ventricular function to stratify risk. Although nuclear cardiology, angiography, and echocardiography are options, echocardiography is used most often worldwide. Our angiographic data cannot be translated directly to echocardiography. The study does show that resting left ventricular function remains a powerful predictor even after considering complications and procedures, but when all other information is available, it adds only modestly to the ability to predict outcome.
Several investigators have pointed out the poor outcomes of MI patients with previous bypass surgery. Most of this has been attributed to the high prevalence of left main and 3-vessel disease, although evidence points to a lower rate of reperfusion of occluded grafts. The most appropriate treatment for such patients also has been controversial.
Percutaneous or surgical revascularization after the acute event is associated with an early hazard, followed by a reduced risk of death over the later period. Studies of elective bypass surgery,21 22 angioplasty,23 and revascularization after MI24 all show this pattern. Thus, the lower mortality rates of survivors of the acute phase should not be taken as definitive evidence that revascularization prolongs survival (also keeping in mind the possibility of selection of a less-sick subgroup), just as the increased mortality rates in the first few days after revascularization should not be taken to indicate that revascularization increases mortality rates overall.
Clinicians often wonder about the role of major cardiovascular complications in estimating future risk, even when patients have recovered from such complications. Among patients who survive to 30 days, stroke appears to be the most serious complication, then heart failure, arrhythmia, and shock. The use of mechanical ventilation, angiography, and balloon pumping were markers of poor outcome, even in patients surviving the acute event. Right-heart catheterization was associated with greater late mortality rates. Whether this represents a marker of risk or a procedure leading to improper treatment has been raised in a recent report,25 and this study cannot resolve the issue.
Our results pertain only to patients who present with ST-segment elevation MI who are eligible for fibrinolysis. Patients in fibrinolytic trials have a lower mortality risk than general MI populations.26 Thus, caution must be urged in extrapolating these results directly into predictions about the general population; the absolute mortality predictions in this model may underestimate the mortality rates seen in practice. Predictive factors may differ for other types of patients with MI. Furthermore, 3 noticeable features are missing from this evaluation: provocative tests for ischemia, serial ECGs, and data about discharge medications. The trial did not assess noninvasive testing; only a minority of patients undergoing thrombolysis for acute MI undergo these tests.27 Furthermore, studies of noninvasive tests generally have found that not having such a test is the most important adverse prognostic factor. Thus, although noninvasive test data would be useful to include, the available data (clinical findings and angiography) represent the information most often available in practice. In contrast, serial ECGs are available for all patients in practice; unfortunately, these data are not available from GUSTO-I. Finally, because doses and duration of discharge medications could not be collected in this large, simple trial, we could not analyze such data. Variation in appropriate use of aspirin, β-blockers, lipid-lowering agents, and ACE inhibitors could lead to substantial differences in mortality rates.
This study provides a method for practitioners to advise patients who have survived the acute phase about their prognosis after MI. Although the predictions are far from perfect for the reasons listed above, the ability to speak more directly about expected outcome may enhance planning and the setting of realistic expectations.
This study was funded by Genentech, Inc, South San Francisco, Calif; Sanofi Pharmaceuticals, Paris, France; ICI Pharmaceuticals, Wilmington, Del; Bayer Corporation, New York, NY; and Ciba-Corning, Medfield, Mass.
- Received August 2, 1999.
- Revision received November 30, 1999.
- Accepted December 13, 1999.
- Copyright © 2000 by American Heart Association
Lee KL, Woodlief LH, Topol EJ, et al. Predictors of 30-day mortality in the era of reperfusion for acute myocardial infarction: results from an international trial of 41 021 patients. Circulation. 1995;91:1659−1668.
Califf RM, White HD, Van de Werf F, et al. One-year results from the Global Utilization of Streptokinase and TPA for Occluded Coronary Arteries (GUSTO-I) trial. Circulation. 1996;94:1233−1238.
Grambsch P, Therneau T. Proportional hazards tests and diagnostics based on weighted residuals. Biometrika. 1994;81:515−526.
Devlin TF, Weeks BJ. Spline functions for logistic regression modeling. In: Proceedings of the 11th Annual SAS Users Group International Conference. Cary, NC: SAS Institute; 1986:646−651.
Harrell FE Jr. Predicting Outcomes: Applied Survival Analysis and Logistic Regression. Charlottesville, Va: University of Virginia Press; 1998.
White HD, Barbash GI, Califf RM, et al. Age and outcome with contemporary thrombolytic therapy: results from the GUSTO-I trial. Circulation. 1996;94:1826−1833.
Newby KH, Thompson T, Stebbins A, et al. Sustained ventricular arrhythmias in patients receiving thrombolytic therapy: incidence and outcomes. Circulation. 1998;98:2567−2573.
Mark DB, Nelson CL, Califf RM, et al. The continuing evolution of therapy for coronary artery disease: initial results from the era of coronary angioplasty. Circulation. 1994;89:2015−2025.
Rogers WJ, Bowlby LJ, Chandra NC, et al. Treatment of myocardial infarction in the United States (1990 to 1993): observations from the National Registry of Myocardial Infarction. Circulation. 1994;90:2103−2114.
Topol EJ, Ellis SG, Cosgrove DM, et al. Analysis of coronary angioplasty practice in the United States with an insurance-claims data base. Circulation. 1993;87:1489−1497.