Author + information
- Received January 26, 2007
- Revision received April 16, 2007
- Accepted April 30, 2007
- Published online October 30, 2007.
- Mandeep Singh, MD⁎,
- Jennifer White, MS†,
- David Hasdai, MD‡,
- Patricia K. Hodgson, BA†,
- Peter B. Berger, MD§,
- Eric J. Topol, MD∥,
- Robert M. Califf, MD† and
- David R. Holmes Jr, MD⁎,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. David R. Holmes, Jr., Division of Cardiovascular Diseases, Department of Internal Medicine, Mayo Clinic, 200 1st Street SW, Rochester, Minnesota 55905.
Objectives This study sought to assess long-term outcome and determine its predictors among 30-day survivors of cardiogenic shock.
Background Patients with cardiogenic shock have high in-hospital and 30-day mortality, but there are little data about those who survive beyond 30 days.
Methods We analyzed baseline, in-hospital, and survival data from patients in the U.S. with ST-segment elevation myocardial infarction (STEMI) and cardiogenic shock enrolled in the GUSTO (Global Utilization of Streptokinase and Tissue-Type Plasminogen Activator for Occluded Coronary Arteries)-I trial and compared them with patients in the same trial who did not have shock.
Results Of 22,883 patients enrolled in the U.S., shock occurred in 1,891 (8.3%); 953 (50.4%) survived 30 days and 527 (27.8%) survived 11 years. Of 20,992 U.S. patients without shock, 20,360 (96.9%) survived 30 days and 14,131 (67.3%) survived 11 years. After the first year, 2% to 4% of patients died each year regardless of whether they had cardiogenic shock. Using Cox proportional hazards models, we were able to predict long-term mortality in all U.S. GUSTO-I 30-day survivors from their baseline demographics and in-hospital complications. The strongest predictors were diabetes mellitus, cardiogenic shock, hypertension, previous myocardial infarction, current smoking, anterior infarct, higher Killip class, higher heart rate, and older age; patients >75 years were at highest risk. Percutaneous revascularization during the index hospitalization was associated with a reduced risk of death.
Conclusions Among patients with cardiogenic shock who survive 30 days after STEMI, annual mortality rates of 2% to 4% approximate those of patients without shock.
Cardiogenic shock is of great interest, from its epidemiology and pathophysiology to therapeutic strategies when it complicates ST-segment elevation myocardial infarction (STEMI) (1–5). Shock accounts for two-thirds of all in-hospital mortality from STEMI, most within the first 30 days, although it occurs in only 4% to 7% of patients (6). Little is known about 30-day survivors of shock.
Predictors of short-term and 1-year outcomes in patients with cardiogenic shock have been described (1,7). Few publications report long-term outcome or predictors, although such data would help optimize treatment strategies (8,9).
The GUSTO (Global Utilization of Streptokinase and Tissue-Type Plasminogen Activator for Occluded Coronary Arteries)-I trial presents an opportunity to evaluate long-term outcome in fibrinolytic-eligible patients with and without cardiogenic shock (10). We describe 11-year outcomes of patients with STEMI with and without cardiogenic shock enrolled in the U.S. in the GUSTO-I trial who survived 30 days after the index event.
The GUSTO-I trial was a multicenter, multinational trial of 41,021 patients with STEMI who presented with ≥1 mm ST-segment elevation in ≥2 limb leads or ≥2 mm in ≥2 contiguous precordial leads 20 min to 6 h after symptom onset between 1990 and 1993 (10). Exclusion criteria were previous stroke, active bleeding, previous treatment with streptokinase or anistreplase, recent trauma or major surgery, previous participation in the trial, or noncompressible vascular punctures. Patients with severe, uncontrolled hypertension (systolic blood pressure ≥180 mm Hg, unresponsive to therapy) had a relative contraindication to enrollment. There was no upper age limit. The primary end point was all-cause 30-day mortality. Patients gave informed consent, and the institutional review board at each hospital approved the protocol.
The GUSTO-I trial did not exclude patients with cardiogenic shock due to any cause and thus did not restrict the definition of shock to left ventricular pump failure. Shock was defined as systolic blood pressure <90 mm Hg for ≥1 h not responsive to fluid administration alone, thought to be secondary to cardiac dysfunction, and associated with signs of hypoperfusion or cardiac index ≤2.2 l/min/mm2and pulmonary capillary wedge pressure >18 mm Hg.
To obtain long-term outcome data from the GUSTO-I trial, we used the National Death Index (NDI), restricting us to following up only patients from the U.S.
For all U.S. patients who survived 30 days, specific data (Social Security number, birth date, gender, first name, last name) were submitted to the NDI and matched against its records to eliminate patients who had died in the interim. When all data elements matched precisely, patients received high probability scores from the NDI, and we considered them nonsurvivors. Occasionally, the NDI assigned slightly lower probabilities to patients who matched on most but not all of the elements (e.g., a minor difference in the spelling of a patient’s last name but all other factors matched exactly). We also considered such patients to be late deaths. Minimum follow-up for the surviving patients was 9.9 years; 95% were followed up for at least 11 years. Events were censored at 11 years for this analysis.
We present stratified Kaplan-Meier estimates for mortality rates by baseline characteristics and index hospitalization events. For continuous variables, suitable cut points were chosen based on the distribution of the data.
Univariable analysis for trends was performed with Cox proportional hazard models. For categorical variables, indicator variables were formed. Univariable models to test categorical variables with more than 2 classes used all indicators less 1 category for the reference group. In the case of myocardial infarction (MI) location, patients were assigned to the hierarchy of anterior, inferior, and other. Results were interpreted as statistically significant when p < 0.05. Overall long-term mortality was depicted with 95% confidence intervals using the Kaplan-Meier method.
Multivariable Cox proportional hazards survival modeling techniques were used to develop a predictive model for long-term mortality risk. All variables listed in Tables 1 and 2⇓⇓were considered in the model. A stepwise selection process was used, augmented by bootstrap techniques to alleviate shortcomings of stepwise selection (11,12). The process consisted of creating 200 bootstrap populations using random selection of study patients, with replacement. Stepwise Cox regression models were run on all bootstrap samples, and factors were confirmed for inclusion in the final model based on the overall frequency of their retention in the models. This method was used in an iterative fashion to refine and confirm selection decisions.
Linearity assumptions were checked, and transformations of continuous variables were created as needed. For all continuous independent variables, a restricted cubic spline was fit and plotted against the dependent variable (outcome of death) (13). Plots of the spline-transformed continuous variables and the log-hazard ratio for death were produced. When these plots showed a linear relationship, no transformations were considered. If the plots depicted nonlinear relationships, simple transformations such as inverse, polynomial, natural logarithm, or linear splines were pursued. The only continuous variable retained in the final multivariable model for long-term survival was age. The transformation that worked best was a linear spline with the cut point at age 55 years; this transformation forms 2 independent continuous variables at the point where the relationship to outcome changes.
Of the 22,883 U.S. patients, 1,891 (8.3%) arrived in or developed shock during the index hospitalization. At 30 days, 20,360 patients (88.9%) without and 953 (50.4%) with shock were alive; their baseline characteristics are presented in Table 3.After a median of 11 years, 14,131 30-day survivors (69.4%) without and 527 (55.2%) with shock remained alive.
Among all 30-day survivors, risk factors for 11-year mortality were as expected: older age, anterior MI, worse Killip class, higher heart rate, and adverse medical history (Table 1). Patients with in-hospital complications were more likely to die in the ensuing 11 years, whereas those receiving percutaneous revascularization during the index hospitalization were less likely to die (Table 2).
For all 30-day survivors, Figure 1shows hazard ratio results of univariable Cox regression modeling of the variables associated with higher long-term mortality, which did not differ between patients with and without shock. Increasing age, lower body weight, comorbid conditions, adverse history, anterior MI, in-hospital complications, and longer hospital stay were all associated with adverse long-term prognosis. Killip class I versus IV at baseline and coronary angioplasty during index hospitalization were associated with long-term survival.
For all patients, regardless of shock status, age >75 years at enrollment was associated with high mortality at 30 days (10.9% nonshock and 68.1% shock patients) and, among 30-day survivors, at 11 years (68.2% and 65.6%, respectively) (Table 4).
Figure 2depicts the Kaplan-Meier curves for 11-year mortality for all 30-day survivors. Beginning in the second year, mortality rates were 2% to 4% per year for all patients regardless of shock status (Fig. 3).
For all 30-day survivors, we used multivariable Cox proportional hazard regression analysis to determine the influence of baseline clinical variables and revascularization on long-term outcome. Table 5shows variables associated with long-term mortality: shock, hypertension, diabetes, current smoking, anterior MI, previous MI, higher heart rate, and worse Killip class. The influence of increasing age had greater impact on outcome for patients ≥55 years at randomization than on those <55 years.
The most important finding of the present study is that the annual mortality of 2% to 4% was similar among U.S. GUSTO-I trial patients with and without cardiogenic shock 2 years to 11 years after STEMI. A second notable finding is that the predictors of long-term outcome were similar regardless of shock status, underscoring the need for aggressive strategies to lower the early hazard associated with shock.
This annual mortality rate of 2% to 4% is similar to 10-year data from the ISIS (International Study of Infarct Survival)-2 trial, with annual mortality of approximately 4% after the first year irrespective of shock status (14). The annual mortality in similar patients enrolled in the GISSI (Gruppo Italiano per lo Studio della Sopravvivenza nell’Infarto Miocardico)-I study was 3.6% (15).
Two studies have published 1-year survival data in shock patients (1,7). The GUSTO-I trial reported survival of 88% among all shock patients worldwide who survived 30 days; survival was 97% in patients without shock (1). The SHOCK (Should We Emergently Revascularize Occluded Coronaries for Cardiogenic Shock) trial of patients with predominantly left ventricular failure reported 1-year survival of 46.7% in the early revascularization group, significantly better than 33.6% in the initial medical stabilization group (7).
Six-year follow-up data are available for the SHOCK and TRACE (Trandolapril Cardiac Evaluation) studies (16,17). The SHOCK study survival was 20% and 33% for, respectively, the early revascularization and initial stabilization groups, whereas TRACE study survival was 55% in nonshock and 12% in shock patients. By contrast, GUSTO-I trial 11-year survival was 67% in nonshock and 28% in shock patients.
The favorable long-term outlook for U.S. GUSTO-I trial patients with shock 2 to 11 years after hospital dismissal was not anticipated. Shock has traditionally been thought to be the result of an irreversible loss of a major amount of left ventricular myocardium, leading to high mortality. However, several observations have failed to support this concept. First, the average ejection fraction in the SHOCK trial was 29% (7). Second, at 1 year, many survivors of cardiogenic shock have good functional status (8). It may be that resolution of severe ischemia and/or neurohormonal abnormalities can explain the potential reversibility of shock. New evidence points to involvement of a systemic inflammatory response, activation of complement, release of cytokines, and expression of inducible nitric oxide synthase in the pathogenesis and outcome of patients presenting with cardiogenic shock (6).
We found the significant predictors of long-term mortality to include expected factors: advancing age, traditional risk factors (including current smoking), and evidence of a large infarct or left ventricular dysfunction. These variables also adversely affected prognosis at 1 year with the exception of current smoking; there is no cogent explanation for this latter observation. Patients who continued to smoke likely fared better in the short term, as shown by less need for additional revascularization procedures (18,19). During longer periods, however, smokers have higher adverse cardiovascular events (19). The other predictors of long-term mortality in GUSTO-I trial 30-day survivors are similar to risk factors in patients presenting without ST-segment elevation and cardiogenic shock (20–22).
The SHOCK investigators noted an age–treatment interaction so that a treatment benefit was apparent only for patients <75 years, among whom 1-year survival was 55% (early revascularization) versus 33% (initial medical stabilization). Current American College of Cardiology/American Heart Association guidelines recommend the use of intra-aortic balloon pumping and early surgical or percutaneous revascularization for suitable shock patients <75 years if it can be performed within 18 h of shock (23). Because the GUSTO-I trial had no exclusion criterion for older age, we were able to observe patients >75 years at enrollment, a subgroup not addressed in contemporary guidelines. Among all U.S. patients >75 years with cardiogenic shock who survived 30 days, the hazard ratio for 11-year mortality was 2.05. Older age itself carried substantial risk, as 68.2% of nonshock and 65.6% of shock patients who survived the first 30 days died by 11 years.
Our results pertain only to patients who presented with STEMI and were treated with fibrinolysis. Thus, caution is urged in extrapolating these results to shock patients who are fibrinolytic-ineligible and to patients with shock after non-STEMI. Given the use of the NDI to determine how many patients from the GUSTO-I trial had died, there were limitations to the follow-up data; information about nonfatal cardiovascular events, other medical conditions and treatments, or the frequency with which patients underwent revascularization after 30 days could not be collected.
Another limitation is the derivation from an older dataset. With advances in technology, pharmacotherapy, and management of patients presenting with STEMI including cardiogenic shock, long-term outcomes from this study may be less relevant today. Although advances in primary angioplasty and adjunctive pharmacotherapy may make our results less applicable, this study provides a benchmark against which current strategies can be compared.
The GUSTO-I study was funded by Bayer, CIBA-Corning Diagnostics, Genentech, ICI Pharmaceuticals, and Sanofi Pharmaceuticals. Dr. Topol was chairman of the study, Dr. Califf was the principal investigator, and Dr. Holmes was a regional director. Each of their institutions received funding through contractual relationships, Duke and Cleveland Clinic with the 5 sponsors and Mayo Clinic with the Cleveland Clinic, which was the Executive Center for the GUSTO-I trial. The GUSTO Steering Committee published a manuscript about patient safety and conflict of interest: J Am Coll Cardiol 1992;19:1123–8.
- Abbreviations and Acronyms
- myocardial infarction
- National Death Index
- ST-segment elevation myocardial infarction
- Received January 26, 2007.
- Revision received April 16, 2007.
- Accepted April 30, 2007.
- American College of Cardiology Foundation
- Berger P.B.,
- Tuttle R.H.,
- Holmes D.R. Jr..,
- et al.
- Chen E.W.,
- Canto J.G.,
- Parsons L.S.,
- et al.
- Dzavik V.,
- Sleeper L.A.,
- Cocke T.P.,
- et al.
- Hasdai D.,
- Holmes D.R. Jr..,
- Califf R.M.,
- et al.,
- GUSTO Investigators
- Hochman J.S.
- Lee L.,
- Erbel R.,
- Brown T.M.,
- Laufer N.,
- Meyer J.,
- O’Neill W.W.
- Stone C.J.,
- Koo C.Y.
- Baigent C.,
- Collins R.,
- Appleby P.,
- Parish S.,
- Sleight P.,
- Peto R.,
- The ISIS-2 (Second International Study of Infarct Survival) collaborative group
- Franzosi M.G.,
- Santoro E.,
- De Vita C.,
- et al.
- Lindholm M.G.,
- Kober L.,
- Boesgaard S.,
- Torp-Pedersen C.,
- Aldershvile J.,
- TRACE Study Group
- Bowers T.R.,
- Terrien E.F.,
- O’Neill W.W.,
- Sachs D.,
- Grines C.L.
- Jacobs A.K.,
- French J.K.,
- Col J.,
- et al.
- Holmes D.R. Jr..,
- Berger P.B.,
- Hochman J.S.,
- et al.
- Mukherjee D.,
- Fang J.,
- Chetcuti S.,
- Moscucci M.,
- Kline-Rogers E.,
- Eagle K.A.
- Antman E.M.